
from pprint import pprint
import numpy as np
import pandas as pd

# Libraries to help with data visualization
import matplotlib.pyplot as plt
import seaborn as sns
import plotly.express as px
from IPython.core.display import display

#preprocessing for model
from sklearn.preprocessing import StandardScaler, MinMaxScaler, RobustScaler, Normalizer
import sklearn.metrics as metrics
from sklearn.model_selection import train_test_split

#tensorflow libraries
import tensorflow as tf

from tensorflow import keras
from tensorflow.keras import Sequential
from tensorflow.keras.layers import Dense, Dropout
from tensorflow.keras.metrics import Precision, Recall, BinaryAccuracy

INFO:tensorflow:Enabling eager execution
INFO:tensorflow:Enabling v2 tensorshape
INFO:tensorflow:Enabling resource variables
INFO:tensorflow:Enabling tensor equality
INFO:tensorflow:Enabling control flow v2


# Utility Functions

#Function to make a summary df to aid EDA
def make_summary_cols(df, cat_threshold=20):
    '''Create a df to summarise the data df based on dtypes, nulls, numeric or not, categorical or not

    df: data threshold
    cat_threshold: number of unique types below which assume categorical

    returns:
    summary_cols: df of summary
    d: dict of column names where keys are numeric_cols, categorical_cols and non_numeric_cols where
    values against the keys contain corresponding column names
    '''
    types = df.dtypes
    types.name = 'col_types'

    nuniques = data.nunique()
    nuniques.name = 'n_uniques'

    nulls = df.isnull().sum()
    nulls.name = 'nulls'

    summary_cols = pd.merge(left=pd.merge(left=nuniques, right=types, left_index=True, right_index=True), right=nulls,\
                            left_index=True, right_index=True).sort_values(by='col_types')
    summary_cols['isnumeric_column'] = summary_cols['col_types'].apply(lambda x: False if x=='object' else True)

    summary_cols['probably_categorical'] = summary_cols['n_uniques'].apply(lambda x: True if x <=cat_threshold \
        else False)

    d = {
        'numeric_cols': list(summary_cols[summary_cols.isnumeric_column==True].index),
        'categorical_cols': list(summary_cols[summary_cols.probably_categorical==True].index),
        'non_numeric_cols': list(summary_cols[summary_cols.isnumeric_column==False].index)
    }

    return summary_cols, d

#Plot histogram and boxplot together
def histogram_boxplot(feature, figsize=(15,10), bins = None):
    """ Boxplot and histogram combined
    feature: 1-d feature array
    figsize: size of fig (default (9,8))
    bins: number of bins (default None / auto)
    """
    f2, (ax_box2, ax_hist2) = plt.subplots(nrows = 2, # Number of rows of the subplot grid= 2
                                           sharex = True, # x-axis will be shared among all subplots
                                           gridspec_kw = {"height_ratios": (.25, .75)}, 
                                           figsize = figsize 
                                           ) # creating the 2 subplots
    sns.boxplot(x=feature, ax=ax_box2, showmeans=True, color='violet') # boxplot will be created and a star will indicate the mean value of the column
    sns.histplot(x=feature, kde=True, ax=ax_hist2, bins=bins,palette="winter") if bins else sns.histplot(x=feature, kde=True, ax=ax_hist2) # For histogram
    ax_hist2.axvline(np.mean(feature), color='green', linestyle='--') # Add mean to the histogram
    ax_hist2.axvline(np.median(feature), color='black', linestyle='-') # Add median to the histogram

#Plot confusion matrix
def make_confusion_matrix(model, y_actual, y_predict=None, labels=[1, 0], cmap='Blues'):
    '''
    model : classifier to predict values of X
    y_actual : ground truth  
    
    '''
    if y_predict is None:
        y_predict = model.predict(X_test)

    cm=metrics.confusion_matrix(y_actual, y_predict, labels=[0, 1])
    df_cm = pd.DataFrame(cm, index = [i for i in ["Actual - No","Actual - Yes"]],
                  columns = [i for i in ['Predicted - No','Predicted - Yes']])
    group_counts = ["{0:0.0f}".format(value) for value in
                cm.flatten()]
    group_percentages = ["{0:.2%}".format(value) for value in
                         cm.flatten()/np.sum(cm)]
    labels = [f"{v1}\n{v2}" for v1, v2 in
              zip(group_counts,group_percentages)]
    labels = np.asarray(labels).reshape(2,2)
    plt.figure(figsize = (10,7))
    sns.heatmap(df_cm, annot=labels,fmt='', cmap=cmap)
    plt.ylabel('True label')
    plt.xlabel('Predicted label')
    
# Function to add data labels to bar plots
def show_values_on_bars(axs):
    def _show_on_single_plot(ax):        
        for p in ax.patches:
            _x = p.get_x() + p.get_width() / 2
            _y = p.get_y() + p.get_height()
            value = '{:.2f}'.format(p.get_height())
            ax.text(_x, _y, value, ha="center") 

    if isinstance(axs, np.ndarray):
        for idx, ax in np.ndenumerate(axs):
            _show_on_single_plot(ax)
    else:
        _show_on_single_plot(axs)


data = pd.read_csv('bank.csv')


data.head()


data.shape

(10000, 14)


data.isnull().sum()

RowNumber          0
CustomerId         0
Surname            0
CreditScore        0
Geography          0
Gender             0
Age                0
Tenure             0
Balance            0
NumOfProducts      0
HasCrCard          0
IsActiveMember     0
EstimatedSalary    0
Exited             0
dtype: int64


data.nunique()

RowNumber          10000
CustomerId         10000
Surname             2932
CreditScore          460
Geography              3
Gender                 2
Age                   70
Tenure                11
Balance             6382
NumOfProducts          4
HasCrCard              2
IsActiveMember         2
EstimatedSalary     9999
Exited                 2
dtype: int64


data.dtypes

RowNumber            int64
CustomerId           int64
Surname             object
CreditScore          int64
Geography           object
Gender              object
Age                  int64
Tenure               int64
Balance            float64
NumOfProducts        int64
HasCrCard            int64
IsActiveMember       int64
EstimatedSalary    float64
Exited               int64
dtype: object


data.describe().T


data.describe(include=['object']).T


# Creating new df dropping unnecessary columns
working_df = data.drop(['RowNumber','CustomerId','Surname'], axis=1)


# create lists of columns by type
summary_df, d = make_summary_cols(working_df)

numeric_cols = d['numeric_cols']
categorical_cols = d['categorical_cols']
non_numeric_cols = d['non_numeric_cols']

numeric_cols.sort()
categorical_cols.sort()
non_numeric_cols.sort()

summary_df


print(f'numeric cols are: \n {numeric_cols}\n')
print(f'categorical cols (defined as <= 20 uniques) are: \n {categorical_cols}\n')
print(f'non numeric cols are: \n {non_numeric_cols}')

numeric cols are: 
 ['Age', 'Balance', 'CreditScore', 'EstimatedSalary', 'Exited', 'HasCrCard', 'IsActiveMember', 'NumOfProducts', 'Tenure']

categorical cols (defined as <= 20 uniques) are: 
 ['Exited', 'Gender', 'Geography', 'HasCrCard', 'IsActiveMember', 'NumOfProducts', 'Tenure']

non numeric cols are: 
 ['Gender', 'Geography']


histogram_boxplot(working_df.CreditScore)


px.box(working_df.CreditScore, orientation='h', height=200)


px.histogram(working_df.CreditScore)


working_df.CreditScore.describe()

count    10000.000000
mean       650.528800
std         96.653299
min        350.000000
25%        584.000000
50%        652.000000
75%        718.000000
max        850.000000
Name: CreditScore, dtype: float64


histogram_boxplot(working_df.Age)


px.box(working_df.Age, orientation='h', height=200)


print(f'No. of rows with age > 62: {working_df[working_df.Age>62].count()[1]}')

No. of rows with age > 62: 359


working_df.Age.describe()

count    10000.000000
mean        38.921800
std         10.487806
min         18.000000
25%         32.000000
50%         37.000000
75%         44.000000
max         92.000000
Name: Age, dtype: float64


histogram_boxplot(working_df.EstimatedSalary)


px.box(working_df.EstimatedSalary,orientation='h',height=200)


working_df.EstimatedSalary.describe()

count     10000.000000
mean     100090.239881
std       57510.492818
min          11.580000
25%       51002.110000
50%      100193.915000
75%      149388.247500
max      199992.480000
Name: EstimatedSalary, dtype: float64


histogram_boxplot(working_df.Balance)


px.box(working_df.Balance, orientation='h', height=200)


#working_df[working_df.Balance == 0].count()
working_df.Balance.value_counts()

0.00         3617
105473.74       2
130170.82       2
72594.00        1
139723.90       1
             ... 
130306.49       1
92895.56        1
132005.77       1
166287.85       1
104001.38       1
Name: Balance, Length: 6382, dtype: int64


working_df.Balance.describe()

count     10000.000000
mean      76485.889288
std       62397.405202
min           0.000000
25%           0.000000
50%       97198.540000
75%      127644.240000
max      250898.090000
Name: Balance, dtype: float64


plt.figure(figsize=(20,5))
zz=sns.countplot(x=working_df.Tenure)
show_values_on_bars(zz)


plt.figure(figsize=(20,5))
zz=sns.countplot(x=working_df.NumOfProducts)
show_values_on_bars(zz)


plt.figure(figsize=(20,5))
zz=sns.countplot(x=working_df.HasCrCard)
show_values_on_bars(zz)


plt.figure(figsize=(20,5))
zz=sns.countplot(x=working_df.IsActiveMember)
show_values_on_bars(zz)


plt.figure(figsize=(20,5))
zz=sns.countplot(x=working_df.Exited)
show_values_on_bars(zz)


plt.figure(figsize=(20,5))
zz=sns.countplot(x=working_df.Geography)
show_values_on_bars(zz)


plt.figure(figsize=(20,5))
zz=sns.countplot(x=working_df.Gender)
show_values_on_bars(zz)


sns.pairplot(working_df,corner=True,diag_kind='kde', hue='Exited')

<seaborn.axisgrid.PairGrid at 0x7fea50c64040>


plt.figure(figsize=(15,15))
sns.heatmap(working_df.loc[:,numeric_cols].corr(), annot=True, fmt='.2f', cmap='coolwarm')

<AxesSubplot:>


working_df.groupby('Exited').mean()


working_df.groupby('Exited').median()


crosstabs = []
for i,col in enumerate(categorical_cols[1:]):
    crosstabs.append(pd.crosstab(working_df.Exited, working_df[col], normalize='columns'))
    display(crosstabs[i])


fig, axes = plt.subplots(6,1, figsize=(25,15))

for i,x in enumerate(crosstabs):
    zz=crosstabs[i].plot(kind='bar',ax=axes[i])
    show_values_on_bars(zz)


working_df[working_df.Balance==0].groupby('Exited').count().iloc[:,1]

Exited
0    3117
1     500
Name: Geography, dtype: int64


temp_df = working_df.copy()
temp_df['BalanceZero'] = temp_df.Balance.apply(lambda x: 1 if x ==0 else 0)

temp_df.groupby('BalanceZero').mean()


del(temp_df)


# Form X and y
X = working_df.drop('Exited', axis=1)
y = working_df.Exited

#Convert categorical non numeric to one hot encoding
X = pd.get_dummies(X, drop_first=True)

#Define random_state for all algorithms that use this
random_state = 314159

X


y.value_counts()

0    7963
1    2037
Name: Exited, dtype: int64


# defined class weights dictionary per keras tutorial
neg, pos = np.bincount(working_df['Exited'])
total = neg + pos
print('class weights:\n    Total: {}\n    Positive: {} ({:.2f}% of total)\n'.format(
    total, pos, 100 * pos / total))

class weights:
    Total: 10000
    Positive: 2037 (20.37% of total)


# define scaler
#https://stackoverflow.com/a/58850139 for ref on which scaler to use
scaler = RobustScaler(quantile_range=(25,75)) #using this as it retains shape of data distribution and also retains outliers 

#in any case RobustScaler yields better results than MinMax or Standard scalers


# Perform train test split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, stratify=y, random_state=random_state)


X_test


# Checking that target split ratio maintained in train and test sets
print(f'% of +ve class in target in full dataset: {working_df[working_df.Exited==1].shape[0]/ working_df.shape[0]}')
print(f'% of +ve class in target in train: {y_train.value_counts()[1]/ len(y_train):0.4f}')
print(f'% of +ve class in target in test: {y_test.value_counts()[1]/ len(y_test):0.4f}')

% of +ve class in target in full dataset: 0.2037
% of +ve class in target in train: 0.2037
% of +ve class in target in test: 0.2037


# Scaling data
# To prevent test data leaking into train, we fit the scaler on train and then transform both train and test on that scaler

X_train = pd.DataFrame(scaler.fit_transform(X_train), columns=X_train.columns)
X_test = pd.DataFrame(scaler.transform(X_test), columns = X_test.columns)



print('Scaled X_train:')
display(X_train.describe().T)

print('\nScaled X_test:')
display(X_test.describe().T)

Scaled X_train:

Scaled X_test:


# Defining metrics - keras tutorial
# I retain these functions here to aid ready reference as needed to tune models
# Modified from source found at: https://www.tensorflow.org/tutorials/structured_data/imbalanced_data

colors = plt.rcParams['axes.prop_cycle'].by_key()['color']

THRESHOLD = 0.5 # threshold used to determine if output is 0 or 1 (since sigmoid outputs a continuous no between 0 and 1)

EPOCHS=100 # baseline epochs for model run

BATCH_SIZE=100 # baseline batchsize

#supress warnings
import tensorflow.python.util.deprecation as deprecation
deprecation._PRINT_DEPRECATION_WARNINGS = False

#metrics for model to compute
METRICS = [
      keras.metrics.TruePositives(name='tp'),
      keras.metrics.FalsePositives(name='fp'),
      keras.metrics.TrueNegatives(name='tn'),
      keras.metrics.FalseNegatives(name='fn'), 
      keras.metrics.BinaryAccuracy(name='accuracy'),
      keras.metrics.Precision(name='precision'),
      keras.metrics.Recall(name='recall'),
      keras.metrics.AUC(name='auc'),
      keras.metrics.AUC(name='prc', curve='PR'), # precision-recall curve
]



#function to define and compile model
def make_model(metrics=METRICS, output_bias=None, dropout=False, learning_rate=1e-3, activation='relu'):
    if output_bias is not None:
        #print('bias being set')
        output_bias = tf.keras.initializers.Constant(output_bias)
    
    model = keras.Sequential()
  
    model.add(keras.layers.Dense(16, activation=activation,input_shape=(X_train.shape[-1],)))
    
    if dropout==True:
        model.add(keras.layers.Dropout(0.5))
    
    model.add(keras.layers.Dense(1, activation='sigmoid', bias_initializer=output_bias))

    model.compile(
        optimizer=keras.optimizers.Adam(learning_rate=learning_rate),
        loss=keras.losses.BinaryCrossentropy(),
        metrics=metrics)

    return model

#variant of above
def make_model2(metrics=METRICS, output_bias=None, dropout=False, learning_rate=1e-3):
    if output_bias is not None:
        #print('bias being set')
        output_bias = tf.keras.initializers.Constant(output_bias)
    
    model = keras.Sequential()
  
    model.add(keras.layers.Dense(64, activation='relu',input_shape=(X_train.shape[-1],)))
    
    if dropout==True:
        model.add(keras.layers.Dropout(0.7))
    
    model.add(keras.layers.Dense(16, activation='relu'))
    
    if dropout==True:
        model.add(keras.layers.Dropout(0.6))

    
    model.add(keras.layers.Dense(1, activation='sigmoid', bias_initializer=output_bias))

    model.compile(
        optimizer=keras.optimizers.Adam(learning_rate=learning_rate),
        loss=keras.losses.BinaryCrossentropy(),
        metrics=metrics)

    return model

#function to plot model metrics (loss, accuracy, recall and precision)
def plot_metrics(history):
  metrics = ['loss', 'accuracy', 'precision', 'recall']
  for n, metric in enumerate(metrics):
    name = metric.replace("_"," ").capitalize()
    plt.subplot(2,2,n+1)
    plt.plot(history.epoch, history.history[metric], color=colors[0], label='Train')
    plt.plot(history.epoch, history.history['val_'+metric],
             color=colors[1], linestyle="--", label='Val')
    plt.xlabel('Epoch')
    plt.ylabel(name)
    if metric == 'loss':
      plt.ylim([0, plt.ylim()[1]])
    elif metric == 'auc':
      plt.ylim([0.8,1])
    else:
      plt.ylim([0,1])

    plt.legend()


%%time
model1 = make_model()
#start with 100 epochs
history1 = model1.fit(X_train, y_train, epochs=EPOCHS, batch_size=BATCH_SIZE, use_multiprocessing=True, validation_split=0.2)

Epoch 1/100
56/56 [==============================] - 2s 13ms/step - loss: 0.6336 - tp: 127.6140 - fp: 469.5965 - tn: 1844.5263 - fn: 456.5088 - accuracy: 0.6479 - precision: 0.2054 - recall: 0.2341 - auc: 0.4712 - prc: 0.2679 - val_loss: 0.5802 - val_tp: 24.0000 - val_fp: 45.0000 - val_tn: 1038.0000 - val_fn: 293.0000 - val_accuracy: 0.7586 - val_precision: 0.3478 - val_recall: 0.0757 - val_auc: 0.4809 - val_prc: 0.2503
Epoch 2/100
56/56 [==============================] - 0s 1ms/step - loss: 0.5304 - tp: 47.6316 - fp: 44.6316 - tn: 2273.7544 - fn: 532.2281 - accuracy: 0.7993 - precision: 0.5217 - recall: 0.0870 - auc: 0.5393 - prc: 0.2914 - val_loss: 0.5439 - val_tp: 8.0000 - val_fp: 21.0000 - val_tn: 1062.0000 - val_fn: 309.0000 - val_accuracy: 0.7643 - val_precision: 0.2759 - val_recall: 0.0252 - val_auc: 0.5511 - val_prc: 0.2747
Epoch 3/100
56/56 [==============================] - 0s 2ms/step - loss: 0.4935 - tp: 22.2281 - fp: 27.3860 - tn: 2309.5088 - fn: 539.1228 - accuracy: 0.8043 - precision: 0.4871 - recall: 0.0434 - auc: 0.5971 - prc: 0.2936 - val_loss: 0.5200 - val_tp: 10.0000 - val_fp: 24.0000 - val_tn: 1059.0000 - val_fn: 307.0000 - val_accuracy: 0.7636 - val_precision: 0.2941 - val_recall: 0.0315 - val_auc: 0.6303 - val_prc: 0.3204
Epoch 4/100
56/56 [==============================] - 0s 2ms/step - loss: 0.4624 - tp: 27.6842 - fp: 36.5439 - tn: 2295.9649 - fn: 538.0526 - accuracy: 0.8027 - precision: 0.4099 - recall: 0.0450 - auc: 0.6883 - prc: 0.3506 - val_loss: 0.5021 - val_tp: 22.0000 - val_fp: 34.0000 - val_tn: 1049.0000 - val_fn: 295.0000 - val_accuracy: 0.7650 - val_precision: 0.3929 - val_recall: 0.0694 - val_auc: 0.6847 - val_prc: 0.3592
Epoch 5/100
56/56 [==============================] - 0s 2ms/step - loss: 0.4506 - tp: 43.7018 - fp: 46.3158 - tn: 2287.9298 - fn: 520.2982 - accuracy: 0.8038 - precision: 0.4776 - recall: 0.0749 - auc: 0.7212 - prc: 0.3865 - val_loss: 0.4890 - val_tp: 30.0000 - val_fp: 36.0000 - val_tn: 1047.0000 - val_fn: 287.0000 - val_accuracy: 0.7693 - val_precision: 0.4545 - val_recall: 0.0946 - val_auc: 0.7138 - val_prc: 0.3925
Epoch 6/100
56/56 [==============================] - 0s 2ms/step - loss: 0.4367 - tp: 70.0351 - fp: 56.4035 - tn: 2272.4035 - fn: 499.4035 - accuracy: 0.8089 - precision: 0.5452 - recall: 0.1157 - auc: 0.7468 - prc: 0.4254 - val_loss: 0.4798 - val_tp: 39.0000 - val_fp: 44.0000 - val_tn: 1039.0000 - val_fn: 278.0000 - val_accuracy: 0.7700 - val_precision: 0.4699 - val_recall: 0.1230 - val_auc: 0.7291 - val_prc: 0.4206
Epoch 7/100
56/56 [==============================] - 0s 2ms/step - loss: 0.4371 - tp: 94.4561 - fp: 60.0702 - tn: 2261.7193 - fn: 482.0000 - accuracy: 0.8112 - precision: 0.6042 - recall: 0.1632 - auc: 0.7483 - prc: 0.4467 - val_loss: 0.4724 - val_tp: 51.0000 - val_fp: 43.0000 - val_tn: 1040.0000 - val_fn: 266.0000 - val_accuracy: 0.7793 - val_precision: 0.5426 - val_recall: 0.1609 - val_auc: 0.7397 - val_prc: 0.4454
Epoch 8/100
56/56 [==============================] - 0s 2ms/step - loss: 0.4223 - tp: 107.1579 - fp: 62.8947 - tn: 2254.6140 - fn: 473.5789 - accuracy: 0.8162 - precision: 0.6227 - recall: 0.1858 - auc: 0.7708 - prc: 0.4721 - val_loss: 0.4663 - val_tp: 59.0000 - val_fp: 45.0000 - val_tn: 1038.0000 - val_fn: 258.0000 - val_accuracy: 0.7836 - val_precision: 0.5673 - val_recall: 0.1861 - val_auc: 0.7475 - val_prc: 0.4674
Epoch 9/100
56/56 [==============================] - 0s 2ms/step - loss: 0.4082 - tp: 115.8246 - fp: 56.9825 - tn: 2275.4386 - fn: 450.0000 - accuracy: 0.8278 - precision: 0.6670 - recall: 0.2160 - auc: 0.7814 - prc: 0.4992 - val_loss: 0.4614 - val_tp: 58.0000 - val_fp: 44.0000 - val_tn: 1039.0000 - val_fn: 259.0000 - val_accuracy: 0.7836 - val_precision: 0.5686 - val_recall: 0.1830 - val_auc: 0.7535 - val_prc: 0.4868
Epoch 10/100
56/56 [==============================] - 0s 2ms/step - loss: 0.4113 - tp: 129.1404 - fp: 55.2807 - tn: 2269.5439 - fn: 444.2807 - accuracy: 0.8261 - precision: 0.6941 - recall: 0.2184 - auc: 0.7841 - prc: 0.5110 - val_loss: 0.4582 - val_tp: 66.0000 - val_fp: 40.0000 - val_tn: 1043.0000 - val_fn: 251.0000 - val_accuracy: 0.7921 - val_precision: 0.6226 - val_recall: 0.2082 - val_auc: 0.7576 - val_prc: 0.4994
Epoch 11/100
56/56 [==============================] - 0s 1ms/step - loss: 0.4025 - tp: 123.0351 - fp: 54.5088 - tn: 2287.4737 - fn: 433.2281 - accuracy: 0.8330 - precision: 0.6895 - recall: 0.2163 - auc: 0.7775 - prc: 0.5084 - val_loss: 0.4539 - val_tp: 71.0000 - val_fp: 42.0000 - val_tn: 1041.0000 - val_fn: 246.0000 - val_accuracy: 0.7943 - val_precision: 0.6283 - val_recall: 0.2240 - val_auc: 0.7628 - val_prc: 0.5123
Epoch 12/100
56/56 [==============================] - 0s 2ms/step - loss: 0.4043 - tp: 158.9649 - fp: 64.7719 - tn: 2258.3333 - fn: 416.1754 - accuracy: 0.8354 - precision: 0.7253 - recall: 0.2833 - auc: 0.7902 - prc: 0.5515 - val_loss: 0.4525 - val_tp: 71.0000 - val_fp: 36.0000 - val_tn: 1047.0000 - val_fn: 246.0000 - val_accuracy: 0.7986 - val_precision: 0.6636 - val_recall: 0.2240 - val_auc: 0.7659 - val_prc: 0.5203
Epoch 13/100
56/56 [==============================] - 0s 2ms/step - loss: 0.4125 - tp: 145.2105 - fp: 63.7895 - tn: 2256.3509 - fn: 432.8947 - accuracy: 0.8231 - precision: 0.6812 - recall: 0.2366 - auc: 0.7825 - prc: 0.5343 - val_loss: 0.4495 - val_tp: 73.0000 - val_fp: 36.0000 - val_tn: 1047.0000 - val_fn: 244.0000 - val_accuracy: 0.8000 - val_precision: 0.6697 - val_recall: 0.2303 - val_auc: 0.7701 - val_prc: 0.5294
Epoch 14/100
56/56 [==============================] - 0s 2ms/step - loss: 0.4010 - tp: 156.6667 - fp: 56.7895 - tn: 2259.8421 - fn: 424.9474 - accuracy: 0.8353 - precision: 0.7553 - recall: 0.2680 - auc: 0.7953 - prc: 0.5633 - val_loss: 0.4480 - val_tp: 73.0000 - val_fp: 36.0000 - val_tn: 1047.0000 - val_fn: 244.0000 - val_accuracy: 0.8000 - val_precision: 0.6697 - val_recall: 0.2303 - val_auc: 0.7732 - val_prc: 0.5373
Epoch 15/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3940 - tp: 158.7895 - fp: 59.4912 - tn: 2271.8772 - fn: 408.0877 - accuracy: 0.8396 - precision: 0.7179 - recall: 0.2809 - auc: 0.7942 - prc: 0.5441 - val_loss: 0.4450 - val_tp: 80.0000 - val_fp: 39.0000 - val_tn: 1044.0000 - val_fn: 237.0000 - val_accuracy: 0.8029 - val_precision: 0.6723 - val_recall: 0.2524 - val_auc: 0.7756 - val_prc: 0.5423
Epoch 16/100
56/56 [==============================] - 0s 2ms/step - loss: 0.4031 - tp: 163.1404 - fp: 64.6140 - tn: 2267.6316 - fn: 402.8596 - accuracy: 0.8374 - precision: 0.7232 - recall: 0.2917 - auc: 0.7848 - prc: 0.5503 - val_loss: 0.4430 - val_tp: 81.0000 - val_fp: 39.0000 - val_tn: 1044.0000 - val_fn: 236.0000 - val_accuracy: 0.8036 - val_precision: 0.6750 - val_recall: 0.2555 - val_auc: 0.7786 - val_prc: 0.5482
Epoch 17/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3843 - tp: 176.2807 - fp: 65.9298 - tn: 2265.3158 - fn: 390.7193 - accuracy: 0.8445 - precision: 0.7469 - recall: 0.3141 - auc: 0.8085 - prc: 0.5909 - val_loss: 0.4405 - val_tp: 82.0000 - val_fp: 43.0000 - val_tn: 1040.0000 - val_fn: 235.0000 - val_accuracy: 0.8014 - val_precision: 0.6560 - val_recall: 0.2587 - val_auc: 0.7804 - val_prc: 0.5533
Epoch 18/100
56/56 [==============================] - 0s 2ms/step - loss: 0.4032 - tp: 177.7018 - fp: 72.4737 - tn: 2237.8246 - fn: 410.2456 - accuracy: 0.8280 - precision: 0.6909 - recall: 0.2913 - auc: 0.7992 - prc: 0.5610 - val_loss: 0.4392 - val_tp: 83.0000 - val_fp: 39.0000 - val_tn: 1044.0000 - val_fn: 234.0000 - val_accuracy: 0.8050 - val_precision: 0.6803 - val_recall: 0.2618 - val_auc: 0.7835 - val_prc: 0.5592
Epoch 19/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3974 - tp: 172.1228 - fp: 58.4035 - tn: 2266.1930 - fn: 401.5263 - accuracy: 0.8389 - precision: 0.7425 - recall: 0.2908 - auc: 0.7959 - prc: 0.5576 - val_loss: 0.4369 - val_tp: 87.0000 - val_fp: 42.0000 - val_tn: 1041.0000 - val_fn: 230.0000 - val_accuracy: 0.8057 - val_precision: 0.6744 - val_recall: 0.2744 - val_auc: 0.7863 - val_prc: 0.5639
Epoch 20/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3838 - tp: 180.1579 - fp: 70.2105 - tn: 2259.1053 - fn: 388.7719 - accuracy: 0.8428 - precision: 0.7186 - recall: 0.3215 - auc: 0.8075 - prc: 0.5807 - val_loss: 0.4351 - val_tp: 87.0000 - val_fp: 42.0000 - val_tn: 1041.0000 - val_fn: 230.0000 - val_accuracy: 0.8057 - val_precision: 0.6744 - val_recall: 0.2744 - val_auc: 0.7886 - val_prc: 0.5698
Epoch 21/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3747 - tp: 198.3158 - fp: 69.0175 - tn: 2258.9649 - fn: 371.9474 - accuracy: 0.8510 - precision: 0.7532 - recall: 0.3572 - auc: 0.8200 - prc: 0.6056 - val_loss: 0.4322 - val_tp: 91.0000 - val_fp: 45.0000 - val_tn: 1038.0000 - val_fn: 226.0000 - val_accuracy: 0.8064 - val_precision: 0.6691 - val_recall: 0.2871 - val_auc: 0.7914 - val_prc: 0.5752
Epoch 22/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3870 - tp: 194.1930 - fp: 71.1754 - tn: 2250.9649 - fn: 381.9123 - accuracy: 0.8412 - precision: 0.7218 - recall: 0.3363 - auc: 0.8103 - prc: 0.5848 - val_loss: 0.4303 - val_tp: 90.0000 - val_fp: 40.0000 - val_tn: 1043.0000 - val_fn: 227.0000 - val_accuracy: 0.8093 - val_precision: 0.6923 - val_recall: 0.2839 - val_auc: 0.7949 - val_prc: 0.5809
Epoch 23/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3828 - tp: 191.9474 - fp: 68.9298 - tn: 2255.9123 - fn: 381.4561 - accuracy: 0.8440 - precision: 0.7197 - recall: 0.3267 - auc: 0.8089 - prc: 0.5786 - val_loss: 0.4280 - val_tp: 92.0000 - val_fp: 40.0000 - val_tn: 1043.0000 - val_fn: 225.0000 - val_accuracy: 0.8107 - val_precision: 0.6970 - val_recall: 0.2902 - val_auc: 0.7974 - val_prc: 0.5870
Epoch 24/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3777 - tp: 205.8246 - fp: 66.0000 - tn: 2252.3860 - fn: 374.0351 - accuracy: 0.8489 - precision: 0.7641 - recall: 0.3645 - auc: 0.8239 - prc: 0.6175 - val_loss: 0.4260 - val_tp: 95.0000 - val_fp: 40.0000 - val_tn: 1043.0000 - val_fn: 222.0000 - val_accuracy: 0.8129 - val_precision: 0.7037 - val_recall: 0.2997 - val_auc: 0.7994 - val_prc: 0.5916
Epoch 25/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3579 - tp: 198.5789 - fp: 59.1228 - tn: 2274.9649 - fn: 365.5789 - accuracy: 0.8598 - precision: 0.7790 - recall: 0.3651 - auc: 0.8309 - prc: 0.6337 - val_loss: 0.4226 - val_tp: 99.0000 - val_fp: 42.0000 - val_tn: 1041.0000 - val_fn: 218.0000 - val_accuracy: 0.8143 - val_precision: 0.7021 - val_recall: 0.3123 - val_auc: 0.8028 - val_prc: 0.5986
Epoch 26/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3743 - tp: 208.1053 - fp: 66.7895 - tn: 2254.6842 - fn: 368.6667 - accuracy: 0.8479 - precision: 0.7492 - recall: 0.3596 - auc: 0.8248 - prc: 0.6136 - val_loss: 0.4214 - val_tp: 99.0000 - val_fp: 40.0000 - val_tn: 1043.0000 - val_fn: 218.0000 - val_accuracy: 0.8157 - val_precision: 0.7122 - val_recall: 0.3123 - val_auc: 0.8052 - val_prc: 0.6029
Epoch 27/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3710 - tp: 213.2456 - fp: 71.6140 - tn: 2246.0351 - fn: 367.3509 - accuracy: 0.8478 - precision: 0.7451 - recall: 0.3706 - auc: 0.8336 - prc: 0.6186 - val_loss: 0.4184 - val_tp: 100.0000 - val_fp: 40.0000 - val_tn: 1043.0000 - val_fn: 217.0000 - val_accuracy: 0.8164 - val_precision: 0.7143 - val_recall: 0.3155 - val_auc: 0.8087 - val_prc: 0.6104
Epoch 28/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3635 - tp: 212.9825 - fp: 66.0351 - tn: 2253.3158 - fn: 365.9123 - accuracy: 0.8507 - precision: 0.7640 - recall: 0.3685 - auc: 0.8387 - prc: 0.6460 - val_loss: 0.4153 - val_tp: 105.0000 - val_fp: 42.0000 - val_tn: 1041.0000 - val_fn: 212.0000 - val_accuracy: 0.8186 - val_precision: 0.7143 - val_recall: 0.3312 - val_auc: 0.8107 - val_prc: 0.6147
Epoch 29/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3658 - tp: 210.2632 - fp: 69.4211 - tn: 2255.1930 - fn: 363.3684 - accuracy: 0.8519 - precision: 0.7465 - recall: 0.3633 - auc: 0.8301 - prc: 0.6159 - val_loss: 0.4125 - val_tp: 108.0000 - val_fp: 42.0000 - val_tn: 1041.0000 - val_fn: 209.0000 - val_accuracy: 0.8207 - val_precision: 0.7200 - val_recall: 0.3407 - val_auc: 0.8143 - val_prc: 0.6211
Epoch 30/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3643 - tp: 236.6491 - fp: 66.8772 - tn: 2239.8596 - fn: 354.8596 - accuracy: 0.8563 - precision: 0.7920 - recall: 0.4170 - auc: 0.8416 - prc: 0.6688 - val_loss: 0.4121 - val_tp: 104.0000 - val_fp: 38.0000 - val_tn: 1045.0000 - val_fn: 213.0000 - val_accuracy: 0.8207 - val_precision: 0.7324 - val_recall: 0.3281 - val_auc: 0.8160 - val_prc: 0.6249
Epoch 31/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3639 - tp: 222.2105 - fp: 67.8947 - tn: 2249.2456 - fn: 358.8947 - accuracy: 0.8528 - precision: 0.7645 - recall: 0.3785 - auc: 0.8356 - prc: 0.6382 - val_loss: 0.4101 - val_tp: 106.0000 - val_fp: 37.0000 - val_tn: 1046.0000 - val_fn: 211.0000 - val_accuracy: 0.8229 - val_precision: 0.7413 - val_recall: 0.3344 - val_auc: 0.8184 - val_prc: 0.6288
Epoch 32/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3597 - tp: 218.9825 - fp: 62.2982 - tn: 2264.7544 - fn: 352.2105 - accuracy: 0.8581 - precision: 0.7871 - recall: 0.3767 - auc: 0.8328 - prc: 0.6419 - val_loss: 0.4067 - val_tp: 112.0000 - val_fp: 39.0000 - val_tn: 1044.0000 - val_fn: 205.0000 - val_accuracy: 0.8257 - val_precision: 0.7417 - val_recall: 0.3533 - val_auc: 0.8210 - val_prc: 0.6329
Epoch 33/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3562 - tp: 218.3509 - fp: 66.0000 - tn: 2263.3509 - fn: 350.5439 - accuracy: 0.8564 - precision: 0.7635 - recall: 0.3860 - auc: 0.8403 - prc: 0.6390 - val_loss: 0.4030 - val_tp: 120.0000 - val_fp: 44.0000 - val_tn: 1039.0000 - val_fn: 197.0000 - val_accuracy: 0.8279 - val_precision: 0.7317 - val_recall: 0.3785 - val_auc: 0.8231 - val_prc: 0.6390
Epoch 34/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3672 - tp: 252.2105 - fp: 78.1228 - tn: 2226.8246 - fn: 341.0877 - accuracy: 0.8495 - precision: 0.7634 - recall: 0.4264 - auc: 0.8477 - prc: 0.6658 - val_loss: 0.4043 - val_tp: 112.0000 - val_fp: 34.0000 - val_tn: 1049.0000 - val_fn: 205.0000 - val_accuracy: 0.8293 - val_precision: 0.7671 - val_recall: 0.3533 - val_auc: 0.8258 - val_prc: 0.6416
Epoch 35/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3573 - tp: 241.5088 - fp: 70.5263 - tn: 2245.9649 - fn: 340.2456 - accuracy: 0.8582 - precision: 0.7784 - recall: 0.4206 - auc: 0.8462 - prc: 0.6613 - val_loss: 0.4018 - val_tp: 117.0000 - val_fp: 39.0000 - val_tn: 1044.0000 - val_fn: 200.0000 - val_accuracy: 0.8293 - val_precision: 0.7500 - val_recall: 0.3691 - val_auc: 0.8272 - val_prc: 0.6453
Epoch 36/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3415 - tp: 227.1053 - fp: 63.2105 - tn: 2276.5965 - fn: 331.3333 - accuracy: 0.8651 - precision: 0.7873 - recall: 0.3967 - auc: 0.8453 - prc: 0.6662 - val_loss: 0.3981 - val_tp: 125.0000 - val_fp: 44.0000 - val_tn: 1039.0000 - val_fn: 192.0000 - val_accuracy: 0.8314 - val_precision: 0.7396 - val_recall: 0.3943 - val_auc: 0.8282 - val_prc: 0.6492
Epoch 37/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3441 - tp: 259.4912 - fp: 75.3684 - tn: 2240.1053 - fn: 323.2807 - accuracy: 0.8610 - precision: 0.7807 - recall: 0.4511 - auc: 0.8620 - prc: 0.6968 - val_loss: 0.3982 - val_tp: 122.0000 - val_fp: 41.0000 - val_tn: 1042.0000 - val_fn: 195.0000 - val_accuracy: 0.8314 - val_precision: 0.7485 - val_recall: 0.3849 - val_auc: 0.8297 - val_prc: 0.6515
Epoch 38/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3424 - tp: 236.2105 - fp: 74.2456 - tn: 2257.0526 - fn: 330.7368 - accuracy: 0.8605 - precision: 0.7575 - recall: 0.4215 - auc: 0.8543 - prc: 0.6658 - val_loss: 0.3947 - val_tp: 130.0000 - val_fp: 44.0000 - val_tn: 1039.0000 - val_fn: 187.0000 - val_accuracy: 0.8350 - val_precision: 0.7471 - val_recall: 0.4101 - val_auc: 0.8308 - val_prc: 0.6553
Epoch 39/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3559 - tp: 247.5614 - fp: 78.1053 - tn: 2242.1930 - fn: 330.3860 - accuracy: 0.8545 - precision: 0.7431 - recall: 0.4289 - auc: 0.8438 - prc: 0.6597 - val_loss: 0.3958 - val_tp: 124.0000 - val_fp: 39.0000 - val_tn: 1044.0000 - val_fn: 193.0000 - val_accuracy: 0.8343 - val_precision: 0.7607 - val_recall: 0.3912 - val_auc: 0.8326 - val_prc: 0.6565
Epoch 40/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3442 - tp: 249.8421 - fp: 72.1228 - tn: 2246.5789 - fn: 329.7018 - accuracy: 0.8596 - precision: 0.7717 - recall: 0.4237 - auc: 0.8580 - prc: 0.6770 - val_loss: 0.3933 - val_tp: 132.0000 - val_fp: 41.0000 - val_tn: 1042.0000 - val_fn: 185.0000 - val_accuracy: 0.8386 - val_precision: 0.7630 - val_recall: 0.4164 - val_auc: 0.8334 - val_prc: 0.6603
Epoch 41/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3338 - tp: 246.8421 - fp: 74.5965 - tn: 2255.7193 - fn: 321.0877 - accuracy: 0.8638 - precision: 0.7560 - recall: 0.4378 - auc: 0.8612 - prc: 0.6815 - val_loss: 0.3911 - val_tp: 136.0000 - val_fp: 45.0000 - val_tn: 1038.0000 - val_fn: 181.0000 - val_accuracy: 0.8386 - val_precision: 0.7514 - val_recall: 0.4290 - val_auc: 0.8336 - val_prc: 0.6618
Epoch 42/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3431 - tp: 249.7895 - fp: 81.5965 - tn: 2236.4912 - fn: 330.3684 - accuracy: 0.8597 - precision: 0.7471 - recall: 0.4378 - auc: 0.8544 - prc: 0.6711 - val_loss: 0.3916 - val_tp: 132.0000 - val_fp: 39.0000 - val_tn: 1044.0000 - val_fn: 185.0000 - val_accuracy: 0.8400 - val_precision: 0.7719 - val_recall: 0.4164 - val_auc: 0.8356 - val_prc: 0.6634
Epoch 43/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3419 - tp: 241.4561 - fp: 71.3509 - tn: 2250.7368 - fn: 334.7018 - accuracy: 0.8619 - precision: 0.7888 - recall: 0.4134 - auc: 0.8555 - prc: 0.6770 - val_loss: 0.3913 - val_tp: 133.0000 - val_fp: 39.0000 - val_tn: 1044.0000 - val_fn: 184.0000 - val_accuracy: 0.8407 - val_precision: 0.7733 - val_recall: 0.4196 - val_auc: 0.8355 - val_prc: 0.6651
Epoch 44/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3511 - tp: 259.5088 - fp: 76.6140 - tn: 2229.0877 - fn: 333.0351 - accuracy: 0.8569 - precision: 0.7799 - recall: 0.4425 - auc: 0.8558 - prc: 0.6887 - val_loss: 0.3911 - val_tp: 133.0000 - val_fp: 39.0000 - val_tn: 1044.0000 - val_fn: 184.0000 - val_accuracy: 0.8407 - val_precision: 0.7733 - val_recall: 0.4196 - val_auc: 0.8364 - val_prc: 0.6656
Epoch 45/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3443 - tp: 240.4386 - fp: 72.9649 - tn: 2256.6316 - fn: 328.2105 - accuracy: 0.8609 - precision: 0.7597 - recall: 0.4157 - auc: 0.8491 - prc: 0.6609 - val_loss: 0.3899 - val_tp: 134.0000 - val_fp: 40.0000 - val_tn: 1043.0000 - val_fn: 183.0000 - val_accuracy: 0.8407 - val_precision: 0.7701 - val_recall: 0.4227 - val_auc: 0.8372 - val_prc: 0.6660
Epoch 46/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3362 - tp: 265.1404 - fp: 73.1930 - tn: 2253.1404 - fn: 306.7719 - accuracy: 0.8663 - precision: 0.7842 - recall: 0.4687 - auc: 0.8627 - prc: 0.7016 - val_loss: 0.3881 - val_tp: 138.0000 - val_fp: 41.0000 - val_tn: 1042.0000 - val_fn: 179.0000 - val_accuracy: 0.8429 - val_precision: 0.7709 - val_recall: 0.4353 - val_auc: 0.8370 - val_prc: 0.6684
Epoch 47/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3295 - tp: 247.1930 - fp: 76.9825 - tn: 2262.2281 - fn: 311.8421 - accuracy: 0.8682 - precision: 0.7523 - recall: 0.4436 - auc: 0.8588 - prc: 0.6685 - val_loss: 0.3869 - val_tp: 138.0000 - val_fp: 45.0000 - val_tn: 1038.0000 - val_fn: 179.0000 - val_accuracy: 0.8400 - val_precision: 0.7541 - val_recall: 0.4353 - val_auc: 0.8375 - val_prc: 0.6698
Epoch 48/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3339 - tp: 263.6842 - fp: 75.0877 - tn: 2251.2632 - fn: 308.2105 - accuracy: 0.8711 - precision: 0.7923 - recall: 0.4774 - auc: 0.8588 - prc: 0.7034 - val_loss: 0.3882 - val_tp: 137.0000 - val_fp: 41.0000 - val_tn: 1042.0000 - val_fn: 180.0000 - val_accuracy: 0.8421 - val_precision: 0.7697 - val_recall: 0.4322 - val_auc: 0.8378 - val_prc: 0.6707
Epoch 49/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3286 - tp: 252.0000 - fp: 77.9298 - tn: 2248.3860 - fn: 319.9298 - accuracy: 0.8642 - precision: 0.7521 - recall: 0.4347 - auc: 0.8625 - prc: 0.6866 - val_loss: 0.3875 - val_tp: 137.0000 - val_fp: 40.0000 - val_tn: 1043.0000 - val_fn: 180.0000 - val_accuracy: 0.8429 - val_precision: 0.7740 - val_recall: 0.4322 - val_auc: 0.8386 - val_prc: 0.6721
Epoch 50/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3285 - tp: 256.0000 - fp: 73.7018 - tn: 2250.1754 - fn: 318.3684 - accuracy: 0.8647 - precision: 0.7753 - recall: 0.4459 - auc: 0.8704 - prc: 0.6973 - val_loss: 0.3864 - val_tp: 137.0000 - val_fp: 42.0000 - val_tn: 1041.0000 - val_fn: 180.0000 - val_accuracy: 0.8414 - val_precision: 0.7654 - val_recall: 0.4322 - val_auc: 0.8390 - val_prc: 0.6719
Epoch 51/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3271 - tp: 261.6491 - fp: 73.4561 - tn: 2251.4211 - fn: 311.7193 - accuracy: 0.8704 - precision: 0.7882 - recall: 0.4613 - auc: 0.8651 - prc: 0.7040 - val_loss: 0.3857 - val_tp: 138.0000 - val_fp: 44.0000 - val_tn: 1039.0000 - val_fn: 179.0000 - val_accuracy: 0.8407 - val_precision: 0.7582 - val_recall: 0.4353 - val_auc: 0.8394 - val_prc: 0.6734
Epoch 52/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3342 - tp: 256.2456 - fp: 76.1228 - tn: 2246.3509 - fn: 319.5263 - accuracy: 0.8618 - precision: 0.7676 - recall: 0.4471 - auc: 0.8640 - prc: 0.6965 - val_loss: 0.3864 - val_tp: 138.0000 - val_fp: 42.0000 - val_tn: 1041.0000 - val_fn: 179.0000 - val_accuracy: 0.8421 - val_precision: 0.7667 - val_recall: 0.4353 - val_auc: 0.8392 - val_prc: 0.6744
Epoch 53/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3339 - tp: 257.2632 - fp: 73.1754 - tn: 2247.0175 - fn: 320.7895 - accuracy: 0.8643 - precision: 0.7778 - recall: 0.4496 - auc: 0.8664 - prc: 0.6893 - val_loss: 0.3857 - val_tp: 138.0000 - val_fp: 42.0000 - val_tn: 1041.0000 - val_fn: 179.0000 - val_accuracy: 0.8421 - val_precision: 0.7667 - val_recall: 0.4353 - val_auc: 0.8388 - val_prc: 0.6742
Epoch 54/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3380 - tp: 267.9298 - fp: 74.8246 - tn: 2238.5614 - fn: 316.9298 - accuracy: 0.8659 - precision: 0.8037 - recall: 0.4552 - auc: 0.8585 - prc: 0.7049 - val_loss: 0.3880 - val_tp: 135.0000 - val_fp: 35.0000 - val_tn: 1048.0000 - val_fn: 182.0000 - val_accuracy: 0.8450 - val_precision: 0.7941 - val_recall: 0.4259 - val_auc: 0.8403 - val_prc: 0.6761
Epoch 55/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3217 - tp: 258.5088 - fp: 72.2105 - tn: 2258.8947 - fn: 308.6316 - accuracy: 0.8718 - precision: 0.7949 - recall: 0.4679 - auc: 0.8724 - prc: 0.7091 - val_loss: 0.3846 - val_tp: 138.0000 - val_fp: 45.0000 - val_tn: 1038.0000 - val_fn: 179.0000 - val_accuracy: 0.8400 - val_precision: 0.7541 - val_recall: 0.4353 - val_auc: 0.8402 - val_prc: 0.6758
Epoch 56/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3366 - tp: 265.6140 - fp: 83.8596 - tn: 2235.0526 - fn: 313.7193 - accuracy: 0.8575 - precision: 0.7566 - recall: 0.4416 - auc: 0.8676 - prc: 0.6932 - val_loss: 0.3852 - val_tp: 138.0000 - val_fp: 41.0000 - val_tn: 1042.0000 - val_fn: 179.0000 - val_accuracy: 0.8429 - val_precision: 0.7709 - val_recall: 0.4353 - val_auc: 0.8403 - val_prc: 0.6769
Epoch 57/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3323 - tp: 256.3684 - fp: 76.4912 - tn: 2245.8947 - fn: 319.4912 - accuracy: 0.8641 - precision: 0.7645 - recall: 0.4415 - auc: 0.8612 - prc: 0.6876 - val_loss: 0.3858 - val_tp: 137.0000 - val_fp: 41.0000 - val_tn: 1042.0000 - val_fn: 180.0000 - val_accuracy: 0.8421 - val_precision: 0.7697 - val_recall: 0.4322 - val_auc: 0.8408 - val_prc: 0.6771
Epoch 58/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3313 - tp: 260.6667 - fp: 74.4211 - tn: 2243.1930 - fn: 319.9649 - accuracy: 0.8643 - precision: 0.7863 - recall: 0.4452 - auc: 0.8679 - prc: 0.6992 - val_loss: 0.3864 - val_tp: 136.0000 - val_fp: 39.0000 - val_tn: 1044.0000 - val_fn: 181.0000 - val_accuracy: 0.8429 - val_precision: 0.7771 - val_recall: 0.4290 - val_auc: 0.8412 - val_prc: 0.6771
Epoch 59/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3227 - tp: 271.6491 - fp: 69.6140 - tn: 2247.8246 - fn: 309.1579 - accuracy: 0.8732 - precision: 0.8185 - recall: 0.4753 - auc: 0.8742 - prc: 0.7208 - val_loss: 0.3849 - val_tp: 136.0000 - val_fp: 42.0000 - val_tn: 1041.0000 - val_fn: 181.0000 - val_accuracy: 0.8407 - val_precision: 0.7640 - val_recall: 0.4290 - val_auc: 0.8410 - val_prc: 0.6780
Epoch 60/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3266 - tp: 262.5789 - fp: 70.5789 - tn: 2254.7368 - fn: 310.3509 - accuracy: 0.8730 - precision: 0.7946 - recall: 0.4543 - auc: 0.8610 - prc: 0.6894 - val_loss: 0.3846 - val_tp: 138.0000 - val_fp: 45.0000 - val_tn: 1038.0000 - val_fn: 179.0000 - val_accuracy: 0.8400 - val_precision: 0.7541 - val_recall: 0.4353 - val_auc: 0.8406 - val_prc: 0.6776
Epoch 61/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3208 - tp: 267.7018 - fp: 77.7018 - tn: 2247.9825 - fn: 304.8596 - accuracy: 0.8682 - precision: 0.7636 - recall: 0.4764 - auc: 0.8707 - prc: 0.7151 - val_loss: 0.3839 - val_tp: 138.0000 - val_fp: 44.0000 - val_tn: 1039.0000 - val_fn: 179.0000 - val_accuracy: 0.8407 - val_precision: 0.7582 - val_recall: 0.4353 - val_auc: 0.8415 - val_prc: 0.6784
Epoch 62/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3260 - tp: 272.6491 - fp: 82.3333 - tn: 2236.1754 - fn: 307.0877 - accuracy: 0.8643 - precision: 0.7661 - recall: 0.4685 - auc: 0.8736 - prc: 0.7073 - val_loss: 0.3843 - val_tp: 138.0000 - val_fp: 43.0000 - val_tn: 1040.0000 - val_fn: 179.0000 - val_accuracy: 0.8414 - val_precision: 0.7624 - val_recall: 0.4353 - val_auc: 0.8416 - val_prc: 0.6785
Epoch 63/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3323 - tp: 269.6491 - fp: 79.5263 - tn: 2236.1579 - fn: 312.9123 - accuracy: 0.8624 - precision: 0.7715 - recall: 0.4632 - auc: 0.8694 - prc: 0.7069 - val_loss: 0.3838 - val_tp: 137.0000 - val_fp: 44.0000 - val_tn: 1039.0000 - val_fn: 180.0000 - val_accuracy: 0.8400 - val_precision: 0.7569 - val_recall: 0.4322 - val_auc: 0.8417 - val_prc: 0.6785
Epoch 64/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3160 - tp: 261.1404 - fp: 75.0000 - tn: 2261.3509 - fn: 300.7544 - accuracy: 0.8717 - precision: 0.7766 - recall: 0.4700 - auc: 0.8739 - prc: 0.7090 - val_loss: 0.3829 - val_tp: 139.0000 - val_fp: 46.0000 - val_tn: 1037.0000 - val_fn: 178.0000 - val_accuracy: 0.8400 - val_precision: 0.7514 - val_recall: 0.4385 - val_auc: 0.8419 - val_prc: 0.6801
Epoch 65/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3289 - tp: 269.2632 - fp: 83.4211 - tn: 2238.5965 - fn: 306.9649 - accuracy: 0.8659 - precision: 0.7566 - recall: 0.4757 - auc: 0.8647 - prc: 0.6997 - val_loss: 0.3834 - val_tp: 138.0000 - val_fp: 46.0000 - val_tn: 1037.0000 - val_fn: 179.0000 - val_accuracy: 0.8393 - val_precision: 0.7500 - val_recall: 0.4353 - val_auc: 0.8425 - val_prc: 0.6799
Epoch 66/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3293 - tp: 262.6491 - fp: 81.8596 - tn: 2242.7544 - fn: 310.9825 - accuracy: 0.8633 - precision: 0.7584 - recall: 0.4696 - auc: 0.8705 - prc: 0.6997 - val_loss: 0.3828 - val_tp: 139.0000 - val_fp: 46.0000 - val_tn: 1037.0000 - val_fn: 178.0000 - val_accuracy: 0.8400 - val_precision: 0.7514 - val_recall: 0.4385 - val_auc: 0.8424 - val_prc: 0.6803
Epoch 67/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3364 - tp: 253.8596 - fp: 83.4386 - tn: 2242.8947 - fn: 318.0526 - accuracy: 0.8574 - precision: 0.7476 - recall: 0.4359 - auc: 0.8633 - prc: 0.6893 - val_loss: 0.3842 - val_tp: 136.0000 - val_fp: 40.0000 - val_tn: 1043.0000 - val_fn: 181.0000 - val_accuracy: 0.8421 - val_precision: 0.7727 - val_recall: 0.4290 - val_auc: 0.8428 - val_prc: 0.6803
Epoch 68/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3304 - tp: 269.3684 - fp: 79.4561 - tn: 2238.6842 - fn: 310.7368 - accuracy: 0.8652 - precision: 0.7793 - recall: 0.4624 - auc: 0.8678 - prc: 0.7030 - val_loss: 0.3825 - val_tp: 139.0000 - val_fp: 45.0000 - val_tn: 1038.0000 - val_fn: 178.0000 - val_accuracy: 0.8407 - val_precision: 0.7554 - val_recall: 0.4385 - val_auc: 0.8430 - val_prc: 0.6814
Epoch 69/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3355 - tp: 250.6491 - fp: 82.9474 - tn: 2246.1930 - fn: 318.4561 - accuracy: 0.8587 - precision: 0.7354 - recall: 0.4275 - auc: 0.8574 - prc: 0.6755 - val_loss: 0.3834 - val_tp: 137.0000 - val_fp: 44.0000 - val_tn: 1039.0000 - val_fn: 180.0000 - val_accuracy: 0.8400 - val_precision: 0.7569 - val_recall: 0.4322 - val_auc: 0.8432 - val_prc: 0.6810
Epoch 70/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3226 - tp: 252.3860 - fp: 75.8596 - tn: 2260.2105 - fn: 309.7895 - accuracy: 0.8678 - precision: 0.7580 - recall: 0.4523 - auc: 0.8669 - prc: 0.6902 - val_loss: 0.3814 - val_tp: 140.0000 - val_fp: 47.0000 - val_tn: 1036.0000 - val_fn: 177.0000 - val_accuracy: 0.8400 - val_precision: 0.7487 - val_recall: 0.4416 - val_auc: 0.8428 - val_prc: 0.6819
Epoch 71/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3315 - tp: 274.7018 - fp: 79.5965 - tn: 2241.1404 - fn: 302.8070 - accuracy: 0.8682 - precision: 0.7823 - recall: 0.4870 - auc: 0.8653 - prc: 0.7117 - val_loss: 0.3818 - val_tp: 139.0000 - val_fp: 48.0000 - val_tn: 1035.0000 - val_fn: 178.0000 - val_accuracy: 0.8386 - val_precision: 0.7433 - val_recall: 0.4385 - val_auc: 0.8427 - val_prc: 0.6811
Epoch 72/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3254 - tp: 260.4737 - fp: 87.6140 - tn: 2243.8772 - fn: 306.2807 - accuracy: 0.8663 - precision: 0.7486 - recall: 0.4649 - auc: 0.8656 - prc: 0.6918 - val_loss: 0.3820 - val_tp: 139.0000 - val_fp: 47.0000 - val_tn: 1036.0000 - val_fn: 178.0000 - val_accuracy: 0.8393 - val_precision: 0.7473 - val_recall: 0.4385 - val_auc: 0.8436 - val_prc: 0.6810
Epoch 73/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3144 - tp: 277.1228 - fp: 76.7544 - tn: 2248.8246 - fn: 295.5439 - accuracy: 0.8732 - precision: 0.7895 - recall: 0.4931 - auc: 0.8784 - prc: 0.7281 - val_loss: 0.3818 - val_tp: 139.0000 - val_fp: 46.0000 - val_tn: 1037.0000 - val_fn: 178.0000 - val_accuracy: 0.8400 - val_precision: 0.7514 - val_recall: 0.4385 - val_auc: 0.8430 - val_prc: 0.6814
Epoch 74/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3311 - tp: 270.2982 - fp: 78.9649 - tn: 2242.9825 - fn: 306.0000 - accuracy: 0.8673 - precision: 0.7768 - recall: 0.4767 - auc: 0.8633 - prc: 0.7047 - val_loss: 0.3821 - val_tp: 137.0000 - val_fp: 45.0000 - val_tn: 1038.0000 - val_fn: 180.0000 - val_accuracy: 0.8393 - val_precision: 0.7527 - val_recall: 0.4322 - val_auc: 0.8434 - val_prc: 0.6826
Epoch 75/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3230 - tp: 264.8421 - fp: 81.5263 - tn: 2246.8070 - fn: 305.0702 - accuracy: 0.8683 - precision: 0.7701 - recall: 0.4786 - auc: 0.8706 - prc: 0.7105 - val_loss: 0.3815 - val_tp: 139.0000 - val_fp: 47.0000 - val_tn: 1036.0000 - val_fn: 178.0000 - val_accuracy: 0.8393 - val_precision: 0.7473 - val_recall: 0.4385 - val_auc: 0.8433 - val_prc: 0.6822
Epoch 76/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3242 - tp: 268.5614 - fp: 79.5965 - tn: 2247.9825 - fn: 302.1053 - accuracy: 0.8690 - precision: 0.7690 - recall: 0.4814 - auc: 0.8705 - prc: 0.7035 - val_loss: 0.3816 - val_tp: 138.0000 - val_fp: 45.0000 - val_tn: 1038.0000 - val_fn: 179.0000 - val_accuracy: 0.8400 - val_precision: 0.7541 - val_recall: 0.4353 - val_auc: 0.8436 - val_prc: 0.6830
Epoch 77/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3202 - tp: 259.7719 - fp: 76.7895 - tn: 2259.1228 - fn: 302.5614 - accuracy: 0.8711 - precision: 0.7753 - recall: 0.4646 - auc: 0.8682 - prc: 0.7005 - val_loss: 0.3806 - val_tp: 141.0000 - val_fp: 50.0000 - val_tn: 1033.0000 - val_fn: 176.0000 - val_accuracy: 0.8386 - val_precision: 0.7382 - val_recall: 0.4448 - val_auc: 0.8437 - val_prc: 0.6828
Epoch 78/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3300 - tp: 268.3860 - fp: 86.6491 - tn: 2236.0000 - fn: 307.2105 - accuracy: 0.8630 - precision: 0.7557 - recall: 0.4633 - auc: 0.8643 - prc: 0.7046 - val_loss: 0.3816 - val_tp: 138.0000 - val_fp: 45.0000 - val_tn: 1038.0000 - val_fn: 179.0000 - val_accuracy: 0.8400 - val_precision: 0.7541 - val_recall: 0.4353 - val_auc: 0.8436 - val_prc: 0.6829
Epoch 79/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3208 - tp: 272.7544 - fp: 76.0175 - tn: 2245.4386 - fn: 304.0351 - accuracy: 0.8733 - precision: 0.7876 - recall: 0.4912 - auc: 0.8712 - prc: 0.7155 - val_loss: 0.3829 - val_tp: 136.0000 - val_fp: 43.0000 - val_tn: 1040.0000 - val_fn: 181.0000 - val_accuracy: 0.8400 - val_precision: 0.7598 - val_recall: 0.4290 - val_auc: 0.8443 - val_prc: 0.6829
Epoch 80/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3212 - tp: 264.7895 - fp: 71.1053 - tn: 2251.2105 - fn: 311.1404 - accuracy: 0.8693 - precision: 0.7993 - recall: 0.4541 - auc: 0.8730 - prc: 0.7116 - val_loss: 0.3813 - val_tp: 137.0000 - val_fp: 45.0000 - val_tn: 1038.0000 - val_fn: 180.0000 - val_accuracy: 0.8393 - val_precision: 0.7527 - val_recall: 0.4322 - val_auc: 0.8445 - val_prc: 0.6839
Epoch 81/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3209 - tp: 262.3509 - fp: 80.9825 - tn: 2241.2632 - fn: 313.6491 - accuracy: 0.8665 - precision: 0.7749 - recall: 0.4578 - auc: 0.8741 - prc: 0.7097 - val_loss: 0.3826 - val_tp: 134.0000 - val_fp: 43.0000 - val_tn: 1040.0000 - val_fn: 183.0000 - val_accuracy: 0.8386 - val_precision: 0.7571 - val_recall: 0.4227 - val_auc: 0.8449 - val_prc: 0.6838
Epoch 82/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3292 - tp: 264.2281 - fp: 72.2105 - tn: 2245.4561 - fn: 316.3509 - accuracy: 0.8662 - precision: 0.7876 - recall: 0.4536 - auc: 0.8674 - prc: 0.7067 - val_loss: 0.3805 - val_tp: 138.0000 - val_fp: 47.0000 - val_tn: 1036.0000 - val_fn: 179.0000 - val_accuracy: 0.8386 - val_precision: 0.7459 - val_recall: 0.4353 - val_auc: 0.8451 - val_prc: 0.6842
Epoch 83/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3156 - tp: 264.2281 - fp: 76.5263 - tn: 2257.2982 - fn: 300.1930 - accuracy: 0.8730 - precision: 0.7737 - recall: 0.4751 - auc: 0.8713 - prc: 0.7079 - val_loss: 0.3791 - val_tp: 145.0000 - val_fp: 50.0000 - val_tn: 1033.0000 - val_fn: 172.0000 - val_accuracy: 0.8414 - val_precision: 0.7436 - val_recall: 0.4574 - val_auc: 0.8442 - val_prc: 0.6841
Epoch 84/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3207 - tp: 259.8772 - fp: 83.8947 - tn: 2243.7018 - fn: 310.7719 - accuracy: 0.8660 - precision: 0.7445 - recall: 0.4509 - auc: 0.8702 - prc: 0.6908 - val_loss: 0.3818 - val_tp: 134.0000 - val_fp: 44.0000 - val_tn: 1039.0000 - val_fn: 183.0000 - val_accuracy: 0.8379 - val_precision: 0.7528 - val_recall: 0.4227 - val_auc: 0.8453 - val_prc: 0.6844
Epoch 85/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3193 - tp: 265.8947 - fp: 70.9825 - tn: 2257.7719 - fn: 303.5965 - accuracy: 0.8718 - precision: 0.7960 - recall: 0.4687 - auc: 0.8732 - prc: 0.7106 - val_loss: 0.3803 - val_tp: 138.0000 - val_fp: 46.0000 - val_tn: 1037.0000 - val_fn: 179.0000 - val_accuracy: 0.8393 - val_precision: 0.7500 - val_recall: 0.4353 - val_auc: 0.8445 - val_prc: 0.6845
Epoch 86/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3310 - tp: 279.9474 - fp: 82.8246 - tn: 2221.4035 - fn: 314.0702 - accuracy: 0.8586 - precision: 0.7633 - recall: 0.4709 - auc: 0.8750 - prc: 0.7144 - val_loss: 0.3822 - val_tp: 135.0000 - val_fp: 44.0000 - val_tn: 1039.0000 - val_fn: 182.0000 - val_accuracy: 0.8386 - val_precision: 0.7542 - val_recall: 0.4259 - val_auc: 0.8451 - val_prc: 0.6842
Epoch 87/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3282 - tp: 267.6842 - fp: 69.7895 - tn: 2255.5965 - fn: 305.1754 - accuracy: 0.8698 - precision: 0.8096 - recall: 0.4532 - auc: 0.8634 - prc: 0.7051 - val_loss: 0.3800 - val_tp: 135.0000 - val_fp: 45.0000 - val_tn: 1038.0000 - val_fn: 182.0000 - val_accuracy: 0.8379 - val_precision: 0.7500 - val_recall: 0.4259 - val_auc: 0.8457 - val_prc: 0.6860
Epoch 88/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3325 - tp: 258.4912 - fp: 73.9474 - tn: 2254.0702 - fn: 311.7368 - accuracy: 0.8641 - precision: 0.7669 - recall: 0.4469 - auc: 0.8626 - prc: 0.6857 - val_loss: 0.3802 - val_tp: 136.0000 - val_fp: 48.0000 - val_tn: 1035.0000 - val_fn: 181.0000 - val_accuracy: 0.8364 - val_precision: 0.7391 - val_recall: 0.4290 - val_auc: 0.8451 - val_prc: 0.6849
Epoch 89/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3174 - tp: 266.5088 - fp: 79.8596 - tn: 2246.2807 - fn: 305.5965 - accuracy: 0.8701 - precision: 0.7629 - recall: 0.4740 - auc: 0.8743 - prc: 0.7043 - val_loss: 0.3799 - val_tp: 135.0000 - val_fp: 47.0000 - val_tn: 1036.0000 - val_fn: 182.0000 - val_accuracy: 0.8364 - val_precision: 0.7418 - val_recall: 0.4259 - val_auc: 0.8451 - val_prc: 0.6852
Epoch 90/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3266 - tp: 263.4737 - fp: 81.1228 - tn: 2244.8947 - fn: 308.7544 - accuracy: 0.8635 - precision: 0.7564 - recall: 0.4636 - auc: 0.8707 - prc: 0.7027 - val_loss: 0.3797 - val_tp: 136.0000 - val_fp: 47.0000 - val_tn: 1036.0000 - val_fn: 181.0000 - val_accuracy: 0.8371 - val_precision: 0.7432 - val_recall: 0.4290 - val_auc: 0.8447 - val_prc: 0.6850
Epoch 91/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3222 - tp: 279.5088 - fp: 79.4386 - tn: 2235.4737 - fn: 303.8246 - accuracy: 0.8691 - precision: 0.7915 - recall: 0.4859 - auc: 0.8750 - prc: 0.7255 - val_loss: 0.3816 - val_tp: 134.0000 - val_fp: 44.0000 - val_tn: 1039.0000 - val_fn: 183.0000 - val_accuracy: 0.8379 - val_precision: 0.7528 - val_recall: 0.4227 - val_auc: 0.8458 - val_prc: 0.6853
Epoch 92/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3083 - tp: 251.3158 - fp: 70.1228 - tn: 2275.3509 - fn: 301.4561 - accuracy: 0.8735 - precision: 0.7714 - recall: 0.4519 - auc: 0.8766 - prc: 0.7041 - val_loss: 0.3786 - val_tp: 145.0000 - val_fp: 51.0000 - val_tn: 1032.0000 - val_fn: 172.0000 - val_accuracy: 0.8407 - val_precision: 0.7398 - val_recall: 0.4574 - val_auc: 0.8446 - val_prc: 0.6852
Epoch 93/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3229 - tp: 265.2632 - fp: 85.4211 - tn: 2246.3509 - fn: 301.2105 - accuracy: 0.8677 - precision: 0.7484 - recall: 0.4646 - auc: 0.8649 - prc: 0.6924 - val_loss: 0.3806 - val_tp: 134.0000 - val_fp: 47.0000 - val_tn: 1036.0000 - val_fn: 183.0000 - val_accuracy: 0.8357 - val_precision: 0.7403 - val_recall: 0.4227 - val_auc: 0.8458 - val_prc: 0.6852
Epoch 94/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3148 - tp: 266.7193 - fp: 70.9123 - tn: 2258.3333 - fn: 302.2807 - accuracy: 0.8747 - precision: 0.8023 - recall: 0.4843 - auc: 0.8773 - prc: 0.7207 - val_loss: 0.3802 - val_tp: 135.0000 - val_fp: 46.0000 - val_tn: 1037.0000 - val_fn: 182.0000 - val_accuracy: 0.8371 - val_precision: 0.7459 - val_recall: 0.4259 - val_auc: 0.8457 - val_prc: 0.6855
Epoch 95/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3130 - tp: 270.4912 - fp: 74.8421 - tn: 2254.7368 - fn: 298.1754 - accuracy: 0.8742 - precision: 0.7865 - recall: 0.4820 - auc: 0.8757 - prc: 0.7176 - val_loss: 0.3796 - val_tp: 137.0000 - val_fp: 50.0000 - val_tn: 1033.0000 - val_fn: 180.0000 - val_accuracy: 0.8357 - val_precision: 0.7326 - val_recall: 0.4322 - val_auc: 0.8455 - val_prc: 0.6852
Epoch 96/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3237 - tp: 271.0351 - fp: 79.8772 - tn: 2243.9474 - fn: 303.3860 - accuracy: 0.8691 - precision: 0.7850 - recall: 0.4800 - auc: 0.8720 - prc: 0.7169 - val_loss: 0.3811 - val_tp: 135.0000 - val_fp: 46.0000 - val_tn: 1037.0000 - val_fn: 182.0000 - val_accuracy: 0.8371 - val_precision: 0.7459 - val_recall: 0.4259 - val_auc: 0.8460 - val_prc: 0.6856
Epoch 97/100
56/56 [==============================] - 0s 1ms/step - loss: 0.3155 - tp: 269.9649 - fp: 79.2982 - tn: 2250.9123 - fn: 298.0702 - accuracy: 0.8719 - precision: 0.7767 - recall: 0.4763 - auc: 0.8743 - prc: 0.7124 - val_loss: 0.3801 - val_tp: 136.0000 - val_fp: 48.0000 - val_tn: 1035.0000 - val_fn: 181.0000 - val_accuracy: 0.8364 - val_precision: 0.7391 - val_recall: 0.4290 - val_auc: 0.8453 - val_prc: 0.6856
Epoch 98/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3236 - tp: 264.7193 - fp: 70.4737 - tn: 2257.7018 - fn: 305.3509 - accuracy: 0.8702 - precision: 0.7949 - recall: 0.4517 - auc: 0.8633 - prc: 0.7039 - val_loss: 0.3798 - val_tp: 134.0000 - val_fp: 49.0000 - val_tn: 1034.0000 - val_fn: 183.0000 - val_accuracy: 0.8343 - val_precision: 0.7322 - val_recall: 0.4227 - val_auc: 0.8455 - val_prc: 0.6857
Epoch 99/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3270 - tp: 263.6316 - fp: 74.2456 - tn: 2256.6667 - fn: 303.7018 - accuracy: 0.8710 - precision: 0.7846 - recall: 0.4696 - auc: 0.8617 - prc: 0.6990 - val_loss: 0.3790 - val_tp: 139.0000 - val_fp: 50.0000 - val_tn: 1033.0000 - val_fn: 178.0000 - val_accuracy: 0.8371 - val_precision: 0.7354 - val_recall: 0.4385 - val_auc: 0.8456 - val_prc: 0.6862
Epoch 100/100
56/56 [==============================] - 0s 2ms/step - loss: 0.3301 - tp: 274.4737 - fp: 84.0877 - tn: 2235.1930 - fn: 304.4912 - accuracy: 0.8655 - precision: 0.7599 - recall: 0.4668 - auc: 0.8638 - prc: 0.6965 - val_loss: 0.3792 - val_tp: 136.0000 - val_fp: 50.0000 - val_tn: 1033.0000 - val_fn: 181.0000 - val_accuracy: 0.8350 - val_precision: 0.7312 - val_recall: 0.4290 - val_auc: 0.8462 - val_prc: 0.6863
CPU times: user 16.3 s, sys: 2.29 s, total: 18.6 s
Wall time: 11 s


model1.summary()

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense (Dense)                (None, 16)                192       
_________________________________________________________________
dense_1 (Dense)              (None, 1)                 17        
=================================================================
Total params: 209
Trainable params: 209
Non-trainable params: 0
_________________________________________________________________


history_df = pd.DataFrame(history1.history)
history_df['epoch']=history1.epoch
display(history_df)
train_acc = history_df.loc[99,'accuracy']
train_recall = history_df.loc[99,'recall']
train_loss = history_df.loc[99,'loss']


results1=model1.evaluate(X_test, y_test.values)
#print(model1.metrics_names)
#print(results1)

results_df = pd.DataFrame(results1, index=model1.metrics_names, columns=['model1'])

results_df

94/94 [==============================] - 1s 1ms/step - loss: 0.3468 - tp: 296.0000 - fp: 103.0000 - tn: 2286.0000 - fn: 315.0000 - accuracy: 0.8607 - precision: 0.7419 - recall: 0.4845 - auc: 0.8518 - prc: 0.6851


plt.figure(figsize=(10,10))
plot_metrics(history1)


# ANN return continuous nos from 0 to 1 as output
# we use a threshold of 0.5 to classify the output as 0 or 1
# the code below converts the true/ false statement into 1/0 by forcing type conversion to int32


y_predict = (model1.predict(X_test) > THRESHOLD).astype('int32')

make_confusion_matrix(model1,y_test,y_predict, cmap='magma_r')


print(f'Model test loss is: {results_df.loc["loss","model1"]:0.4f}, train loss is {train_loss:0.4f}')
print(f'Model test accuracy is: {results_df.loc["accuracy","model1"]:0.4f}, train accuracy is {train_acc:0.4f}')
print(f'Model test recall is: {results_df.loc["recall","model1"]:0.4f}, train recall is {train_recall:0.4f}')

Model test loss is: 0.3468, train loss is 0.3211
Model test accuracy is: 0.8607, train accuracy is 0.8689
Model test recall is: 0.4845, train recall is 0.4788


# Source: https://www.tensorflow.org/tutorials/structured_data/imbalanced_data


# Scaling by total/2 helps keep the loss to a similar magnitude.
# The sum of the weights of all examples stays the same.
weight_for_0 = (1 / neg)*(total)/2.0 
weight_for_1 = (1 / pos)*(total)/2.0

class_weight = {0: weight_for_0, 1: weight_for_1}

print('Weight for class 0: {:.2f}'.format(weight_for_0))
print('Weight for class 1: {:.2f}'.format(weight_for_1))

Weight for class 0: 0.63
Weight for class 1: 2.45


%%time 

model2 = make_model(dropout=True, learning_rate=0.001)

history2 = model2.fit(X_train, y_train, epochs=EPOCHS+100, batch_size=BATCH_SIZE, use_multiprocessing=True, validation_split=0.2, class_weight=class_weight, verbose=0) 
#setting verbose 0 to avoid cluttering the notebook

CPU times: user 29.5 s, sys: 4.28 s, total: 33.7 s
Wall time: 18.5 s


model2.summary()

Model: "sequential_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_2 (Dense)              (None, 16)                192       
_________________________________________________________________
dropout (Dropout)            (None, 16)                0         
_________________________________________________________________
dense_3 (Dense)              (None, 1)                 17        
=================================================================
Total params: 209
Trainable params: 209
Non-trainable params: 0
_________________________________________________________________


history_df = pd.DataFrame(history2.history)
history_df['epoch']=history2.epoch
display(history_df)
train_acc = history_df.loc[199,'accuracy']
train_recall = history_df.loc[199,'recall']
train_loss = history_df.loc[199,'loss']


results_df = results_df['model1']

results2=model2.evaluate(X_test, y_test.values)

temp_df = pd.DataFrame(results2, index=model2.metrics_names, columns=['model2'])

results_df = pd.merge(results_df, temp_df, left_index=True, right_index=True)

results_df

94/94 [==============================] - 1s 1ms/step - loss: 0.4705 - tp: 440.0000 - fp: 452.0000 - tn: 1937.0000 - fn: 171.0000 - accuracy: 0.7923 - precision: 0.4933 - recall: 0.7201 - auc: 0.8495 - prc: 0.6703


plt.figure(figsize=(10,10))
plot_metrics(history2)


y_predict = (model2.predict(X_test) > THRESHOLD).astype('int32')
make_confusion_matrix(model2,y_test, y_predict, cmap='coolwarm')


print(f'Model test loss is: {results_df.loc["loss","model2"]:0.4f}, train loss is {train_loss:0.4f}')
print(f'Model test accuracy is: {results_df.loc["accuracy","model2"]:0.4f}, train accuracy is {train_acc:0.4f}')
print(f'Model test recall is: {results_df.loc["recall","model2"]:0.4f}, train recall is {train_recall:0.4f}')

Model test loss is: 0.4705, train loss is 0.5109
Model test accuracy is: 0.7923, train accuracy is 0.7862
Model test recall is: 0.7201, train recall is 0.6826


%%time
model3 = make_model(dropout=True, learning_rate=0.001, activation='tanh')

history3 = model3.fit(X_train, y_train, epochs=EPOCHS+100, batch_size=BATCH_SIZE, use_multiprocessing=True, validation_split=0.2, class_weight=class_weight, verbose=0)

CPU times: user 27.8 s, sys: 4.26 s, total: 32 s
Wall time: 16.8 s


results_df = results_df.loc[:,['model1','model2']]

results3=model3.evaluate(X_test, y_test.values)

temp_df = pd.DataFrame(results3, index=model3.metrics_names, columns=['model2-tanh'])

results_df = pd.merge(results_df, temp_df, left_index=True, right_index=True)

results_df

94/94 [==============================] - 1s 1ms/step - loss: 0.4949 - tp: 433.0000 - fp: 522.0000 - tn: 1867.0000 - fn: 178.0000 - accuracy: 0.7667 - precision: 0.4534 - recall: 0.7087 - auc: 0.8243 - prc: 0.5848


print(f'Model test loss is: {results_df.loc["loss","model2-tanh"]:0.4f}')
print(f'Model test accuracy is: {results_df.loc["accuracy","model2-tanh"]:0.4f}')
print(f'Model test recall is: {results_df.loc["recall","model2-tanh"]:0.4f}')

Model test loss is: 0.4949
Model test accuracy is: 0.7667
Model test recall is: 0.7087


%%time
model4 = make_model2(dropout=True, learning_rate=0.005, output_bias = np.log(pos/neg)) 

history4 = model4.fit(X_train, y_train, epochs=EPOCHS+100, batch_size=1000, use_multiprocessing=True, validation_split=0.2, class_weight=class_weight, verbose=0)
#setting verbose=0 to avoid cluttering the notebook

CPU times: user 14.4 s, sys: 4.57 s, total: 18.9 s
Wall time: 7.95 s


model4.summary()

Model: "sequential_6"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_15 (Dense)             (None, 64)                768       
_________________________________________________________________
dropout_8 (Dropout)          (None, 64)                0         
_________________________________________________________________
dense_16 (Dense)             (None, 16)                1040      
_________________________________________________________________
dropout_9 (Dropout)          (None, 16)                0         
_________________________________________________________________
dense_17 (Dense)             (None, 1)                 17        
=================================================================
Total params: 1,825
Trainable params: 1,825
Non-trainable params: 0
_________________________________________________________________


history_df = pd.DataFrame(history4.history)
history_df['epoch']=history4.epoch
display(history_df)
train_acc = history_df.loc[199,'accuracy']
train_recall = history_df.loc[199,'recall']
train_loss = history_df.loc[199,'loss']


results_df = results_df.loc[:,['model1','model2','model2-tanh']]

results4=model4.evaluate(X_test, y_test.values)

temp_df = pd.DataFrame(results4, index=model3.metrics_names, columns=['model4'])

results_df = pd.merge(results_df, temp_df, left_index=True, right_index=True)

results_df

94/94 [==============================] - 0s 914us/step - loss: 0.4398 - tp: 435.0000 - fp: 411.0000 - tn: 1978.0000 - fn: 176.0000 - accuracy: 0.8043 - precision: 0.5142 - recall: 0.7119 - auc: 0.8560 - prc: 0.6875


plt.figure(figsize=(10,10))
plot_metrics(history4)


y_predict = (model4.predict(X_test) > THRESHOLD).astype('int32')
make_confusion_matrix(model4,y_test, y_predict, cmap='jet')


print(f'Model test loss is: {results_df.loc["loss","model4"]:0.4f}, train loss is {train_loss:0.4f}')
print(f'Model test accuracy is: {results_df.loc["accuracy","model4"]:0.4f}, train accuracy is {train_acc:0.4f}')
print(f'Model test recall is: {results_df.loc["recall","model4"]:0.4f}, train recall is {train_recall:0.4f}')

Model test loss is: 0.4398, train loss is 0.4725
Model test accuracy is: 0.8043, train accuracy is 0.8048
Model test recall is: 0.7119, train recall is 0.7250


f1=2*results_df.loc['recall']*results_df.loc['precision']/ (results_df.loc['recall']+results_df.loc['precision'])
print('f1 scores: \n')
pprint(f1)

f1 scores: 

model1         0.586139
model2         0.585496
model2-tanh    0.553001
model4         0.597117
dtype: float64


import xgboost as xgb
from xgboost import XGBClassifier

from sklearn.model_selection import RandomizedSearchCV
from sklearn.metrics import plot_confusion_matrix, accuracy_score, recall_score, make_scorer

#xgb.config_context()


params= {
        'n_estimators': np.arange(10,150,10),
        'scale_pos_weight': [(neg/pos)], 
        'learning_rate': [0.05,0.1,0.2,0.3], 
        'gamma': np.arange(0,150,10),
        'subsample': [0.5,0.6,0.7,0.8,0.9,1],
        'colsample_bytree': [0.1,0.2,0.3,0.5,0.7,1],
        'max_depth': [3,5,6,8,10,12,15],
        'tree_method': ['hist'], 
        'min_child_weight': [0,3,5,7,8,9,10],
        'colsample_bylevel': [0.5,0.7,1]
    }


%%time
modelx = XGBClassifier()
scorer = metrics.make_scorer(metrics.recall_score)
cvobj = RandomizedSearchCV(estimator=modelx, param_distributions=params, n_iter=50, \
                            scoring=scorer, cv=5, random_state=random_state)


cvobj.fit(X_train, y_train)

CPU times: user 7min 24s, sys: 22 s, total: 7min 46s
Wall time: 31.9 s

RandomizedSearchCV(cv=5,
                   estimator=XGBClassifier(base_score=None, booster=None,
                                           colsample_bylevel=None,
                                           colsample_bynode=None,
                                           colsample_bytree=None, gamma=None,
                                           gpu_id=None, importance_type='gain',
                                           interaction_constraints=None,
                                           learning_rate=None,
                                           max_delta_step=None, max_depth=None,
                                           min_child_weight=None, missing=nan,
                                           monotone_constraints=None,
                                           n_estimators=100,...
                                        'gamma': array([  0,  10,  20,  30,  40,  50,  60,  70,  80,  90, 100, 110, 120,
       130, 140]),
                                        'learning_rate': [0.05, 0.1, 0.2, 0.3],
                                        'max_depth': [3, 5, 6, 8, 10, 12, 15],
                                        'min_child_weight': [0, 3, 5, 7, 8, 9,
                                                             10],
                                        'n_estimators': array([ 10,  20,  30,  40,  50,  60,  70,  80,  90, 100, 110, 120, 130,
       140]),
                                        'scale_pos_weight': [3.9091801669121256],
                                        'subsample': [0.5, 0.6, 0.7, 0.8, 0.9,
                                                      1],
                                        'tree_method': ['hist']},
                   random_state=314159, scoring=make_scorer(recall_score))


cvobj.best_estimator_

XGBClassifier(base_score=0.5, booster='gbtree', colsample_bylevel=1,
              colsample_bynode=1, colsample_bytree=1, gamma=40, gpu_id=-1,
              importance_type='gain', interaction_constraints='',
              learning_rate=0.2, max_delta_step=0, max_depth=6,
              min_child_weight=3, missing=nan, monotone_constraints='()',
              n_estimators=60, n_jobs=0, num_parallel_tree=1, random_state=0,
              reg_alpha=0, reg_lambda=1, scale_pos_weight=3.9091801669121256,
              subsample=0.6, tree_method='hist', validate_parameters=1,
              verbosity=None)


yypred=cvobj.predict(X_test)


make_confusion_matrix(cvobj, y_test, cmap='Greens')

print(f'accuracy: {accuracy_score(y_test, yypred)}, recall: {recall_score(y_test, yypred)}')

accuracy: 0.792, recall: 0.7430441898527005

	RowNumber	CustomerId	Surname	CreditScore	Geography	Gender	Age	Tenure	Balance	NumOfProducts	HasCrCard	IsActiveMember	EstimatedSalary	Exited
0	1	15634602	Hargrave	619	France	Female	42	2	0.00	1	1	1	101348.88	1
1	2	15647311	Hill	608	Spain	Female	41	1	83807.86	1	0	1	112542.58	0
2	3	15619304	Onio	502	France	Female	42	8	159660.80	3	1	0	113931.57	1
3	4	15701354	Boni	699	France	Female	39	1	0.00	2	0	0	93826.63	0
4	5	15737888	Mitchell	850	Spain	Female	43	2	125510.82	1	1	1	79084.10	0

	CreditScore	Age	Tenure	Balance	NumOfProducts	HasCrCard	IsActiveMember	EstimatedSalary
Exited
0	651.853196	37.408389	5.033279	72745.296779	1.544267	0.707146	0.554565	99738.391772
1	645.351497	44.837997	4.932744	91108.539337	1.475209	0.699067	0.360825	101465.677531

Gender	Female	Male
Exited
0	0.749285	0.835441
1	0.250715	0.164559

Geography	France	Germany	Spain
Exited
0	0.838452	0.675568	0.833266
1	0.161548	0.324432	0.166734

HasCrCard	0	1
Exited
0	0.791851	0.798157
1	0.208149	0.201843

Pre Work¶

Load Libraries and Define Custom Functions¶

1. Load and Inspect Data¶

2.A EDA - Univariate Analysis¶

2.A (i) Credit Score¶

2.A (ii) Age¶

2.A (iii) Estimated Salary¶

2.A (iv) Balance¶

2.A (v) Tenure¶

2.A (vi) NumOfProducts¶

2.A (vii) HasCrCard¶

2.A (viii) IsActiveMember¶

2.A (ix) Exited (Target Variable)¶

2.A (x) Geography¶

2.A (xi) Gender¶

2.B Bivariate Analysis¶

Examining bivariate relationships - vs. Exited¶

Bivariate vs. Balance¶

3. Data Preprocessing¶

4. Modelling¶

Function and Constants needed for section¶

4.A Base model - only one hidden layer and few nodes¶

4.B Refining the ANN Model - using class weights and more epochs¶

4.C Using Tanh as activation for the hidden layers¶

4.D Trying more complex model -¶

Final Conclusions and Business Recommendations¶

Appendix: Using XGBoost to check ANN performance¶

	count	mean	std	min	25%	50%	75%	max
RowNumber	10000.0	5.000500e+03	2886.895680	1.00	2500.75	5.000500e+03	7.500250e+03	10000.00
CustomerId	10000.0	1.569094e+07	71936.186123	15565701.00	15628528.25	1.569074e+07	1.575323e+07	15815690.00
CreditScore	10000.0	6.505288e+02	96.653299	350.00	584.00	6.520000e+02	7.180000e+02	850.00
Age	10000.0	3.892180e+01	10.487806	18.00	32.00	3.700000e+01	4.400000e+01	92.00
Tenure	10000.0	5.012800e+00	2.892174	0.00	3.00	5.000000e+00	7.000000e+00	10.00
Balance	10000.0	7.648589e+04	62397.405202	0.00	0.00	9.719854e+04	1.276442e+05	250898.09
NumOfProducts	10000.0	1.530200e+00	0.581654	1.00	1.00	1.000000e+00	2.000000e+00	4.00
HasCrCard	10000.0	7.055000e-01	0.455840	0.00	0.00	1.000000e+00	1.000000e+00	1.00
IsActiveMember	10000.0	5.151000e-01	0.499797	0.00	0.00	1.000000e+00	1.000000e+00	1.00
EstimatedSalary	10000.0	1.000902e+05	57510.492818	11.58	51002.11	1.001939e+05	1.493882e+05	199992.48
Exited	10000.0	2.037000e-01	0.402769	0.00	0.00	0.000000e+00	0.000000e+00	1.00

	count	unique	top	freq
Surname	10000	2932	Smith	32
Geography	10000	3	France	5014
Gender	10000	2	Male	5457

	n_uniques	col_types	isnumeric_column	probably_categorical
CreditScore	460	int64	True	False
Age	70	int64	True	False
Tenure	11	int64	True	True
NumOfProducts	4	int64	True	True
HasCrCard	2	int64	True	True
IsActiveMember	2	int64	True	True
Exited	2	int64	True	True
Balance	6382	float64	True	False
EstimatedSalary	9999	float64	True	False
Geography	3	object	False	True
Gender	2	object	False	True

	CreditScore	Age	Tenure	Balance	NumOfProducts	HasCrCard	IsActiveMember	EstimatedSalary
Exited
0	653	36	5	92072.68	2	1	1	99645.04
1	646	45	5	109349.29	1	1	0	102460.84

NumOfProducts	1	2	3	4
Exited
0	0.722856	0.924183	0.172932	0.0
1	0.277144	0.075817	0.827068	1.0

Tenure	0	1	2	3	4	5	6	7	8	9	10
Exited
0	0.769976	0.775845	0.808206	0.7889	0.794742	0.793478	0.797311	0.827821	0.807805	0.783537	0.793878
1	0.230024	0.224155	0.191794	0.2111	0.205258	0.206522	0.202689	0.172179	0.192195	0.216463	0.206122

	CreditScore	Age	Tenure	Balance	NumOfProducts	HasCrCard	IsActiveMember	EstimatedSalary	Exited
BalanceZero
0	651.138493	39.197713	4.979633	119827.493793	1.386025	0.699201	0.513552	100717.352956	0.240796
1	649.452861	38.434891	5.071330	0.000000	1.784628	0.716616	0.517832	98983.559549	0.138236

	CreditScore	Age	Tenure	Balance	NumOfProducts	HasCrCard	IsActiveMember	EstimatedSalary	Geography_Germany	Geography_Spain	Gender_Male
2441	735	29	10	0.00	2	1	1	95025.27	0	1	1
2786	511	40	9	124401.60	1	1	0	198814.24	1	0	0
2375	815	39	6	0.00	1	1	1	85167.88	0	1	0
3566	746	25	3	104833.79	1	0	0	71911.30	0	1	0
7616	610	27	4	87262.40	2	1	0	182720.07	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...
2513	666	39	10	0.00	2	1	0	102999.33	0	0	1
8474	721	33	4	72535.45	1	1	1	103931.49	0	1	0
4528	714	31	6	152926.60	1	1	1	50899.91	0	1	0
1732	735	49	5	121973.28	1	1	0	148804.36	0	0	1
5914	754	27	7	117578.35	2	0	1	87908.01	1	0	1

	count	mean	std	min	25%	50%	75%	max
CreditScore	7000.0	-0.011733	0.719086	-2.253731	-0.514925	0.000000e+00	0.485075	1.477612
Age	7000.0	0.168060	0.881617	-1.583333	-0.416667	0.000000e+00	0.583333	4.583333
Tenure	7000.0	-0.001743	0.578801	-1.000000	-0.600000	0.000000e+00	0.400000	1.000000
Balance	7000.0	-0.164601	0.490244	-0.761774	-0.761774	-5.722120e-17	0.238226	1.211395
NumOfProducts	7000.0	0.526571	0.579087	0.000000	0.000000	0.000000e+00	1.000000	3.000000
HasCrCard	7000.0	-0.293000	0.455171	-1.000000	-1.000000	0.000000e+00	0.000000	0.000000
IsActiveMember	7000.0	-0.480714	0.499664	-1.000000	-1.000000	0.000000e+00	0.000000	0.000000
EstimatedSalary	7000.0	0.001752	0.589707	-1.028401	-0.494425	7.470153e-17	0.505575	1.024988
Geography_Germany	7000.0	0.251571	0.433947	0.000000	0.000000	0.000000e+00	1.000000	1.000000
Geography_Spain	7000.0	0.251571	0.433947	0.000000	0.000000	0.000000e+00	1.000000	1.000000
Gender_Male	7000.0	-0.455571	0.498058	-1.000000	-1.000000	0.000000e+00	0.000000	0.000000

	count	mean	std	min	25%	50%	75%	max
CreditScore	3000.0	-0.009219	0.726535	-2.253731	-0.507463	-0.007463	0.501866	1.477612
Age	3000.0	0.141694	0.855769	-1.583333	-0.416667	0.000000	0.500000	4.583333
Tenure	3000.0	0.012600	0.577552	-1.000000	-0.400000	0.000000	0.600000	1.000000
Balance	3000.0	-0.150121	0.491760	-0.761774	-0.761774	0.007951	0.250175	1.113007
NumOfProducts	3000.0	0.538667	0.587611	0.000000	0.000000	0.000000	1.000000	3.000000
HasCrCard	3000.0	-0.298000	0.457456	-1.000000	-1.000000	0.000000	0.000000	0.000000
IsActiveMember	3000.0	-0.494667	0.500055	-1.000000	-1.000000	0.000000	0.000000	0.000000
EstimatedSalary	3000.0	-0.006758	0.592448	-1.027595	-0.520785	0.002163	0.504898	1.024338
Geography_Germany	3000.0	0.249333	0.432699	0.000000	0.000000	0.000000	0.000000	1.000000
Geography_Spain	3000.0	0.238667	0.426340	0.000000	0.000000	0.000000	0.000000	1.000000
Gender_Male	3000.0	-0.451333	0.497709	-1.000000	-1.000000	0.000000	0.000000	0.000000

IsActiveMember	0	1
Exited
0	0.731491	0.857309
1	0.268509	0.142691

	loss	tp	fp	tn	fn	accuracy	precision	recall	auc	prc	...	val_tp	val_fp	val_tn	val_fn	val_accuracy	val_precision	val_recall	val_auc	val_prc	epoch
0	0.596195	201.0	663.0	3828.0	908.0	0.719464	0.232639	0.181244	0.482692	0.254282	...	24.0	45.0	1038.0	293.0	0.758571	0.347826	0.075710	0.480901	0.250323	0
1	0.520762	78.0	83.0	4408.0	1031.0	0.801071	0.484472	0.070334	0.545000	0.270261	...	8.0	21.0	1062.0	309.0	0.764286	0.275862	0.025237	0.551085	0.274663	1
2	0.487728	41.0	54.0	4437.0	1068.0	0.799643	0.431579	0.036970	0.625671	0.305788	...	10.0	24.0	1059.0	307.0	0.763571	0.294118	0.031546	0.630322	0.320409	2
3	0.465317	62.0	78.0	4413.0	1047.0	0.799107	0.442857	0.055906	0.695019	0.359460	...	22.0	34.0	1049.0	295.0	0.765000	0.392857	0.069401	0.684663	0.359158	3
4	0.449409	93.0	89.0	4402.0	1016.0	0.802679	0.510989	0.083859	0.729777	0.399000	...	30.0	36.0	1047.0	287.0	0.769286	0.454545	0.094637	0.713792	0.392455	4
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
95	0.321874	516.0	149.0	4342.0	593.0	0.867500	0.775940	0.465284	0.871937	0.709156	...	135.0	46.0	1037.0	182.0	0.837143	0.745856	0.425867	0.846033	0.685592	95
96	0.321986	527.0	160.0	4331.0	582.0	0.867500	0.767103	0.475203	0.871822	0.709242	...	136.0	48.0	1035.0	181.0	0.836429	0.739130	0.429022	0.845267	0.685617	96
97	0.321585	513.0	143.0	4348.0	596.0	0.868036	0.782012	0.462579	0.872021	0.710075	...	134.0	49.0	1034.0	183.0	0.834286	0.732240	0.422713	0.845467	0.685714	97
98	0.321287	515.0	140.0	4351.0	594.0	0.868929	0.786260	0.464382	0.872737	0.710313	...	139.0	50.0	1033.0	178.0	0.837143	0.735450	0.438486	0.845601	0.686231	98
99	0.321058	531.0	156.0	4335.0	578.0	0.868929	0.772926	0.478810	0.872557	0.710724	...	136.0	50.0	1033.0	181.0	0.835000	0.731183	0.429022	0.846211	0.686295	99

	model1
loss	0.346845
tp	296.000000
fp	103.000000
tn	2286.000000
fn	315.000000
accuracy	0.860667
precision	0.741855
recall	0.484452
auc	0.851803
prc	0.685067

	loss	tp	fp	tn	fn	accuracy	precision	recall	auc	prc	...	val_tp	val_fp	val_tn	val_fn	val_accuracy	val_precision	val_recall	val_auc	val_prc	epoch
0	0.869449	351.0	201.0	6679.0	1369.0	0.817442	0.635870	0.204070	0.674824	0.433570	...	0.0	0.0	1083.0	317.0	0.773571	0.000000	0.000000	0.564945	0.293262	0
1	0.797432	176.0	356.0	4135.0	933.0	0.769821	0.330827	0.158702	0.523980	0.256454	...	2.0	1.0	1082.0	315.0	0.774286	0.666667	0.006309	0.622692	0.345243	1
2	0.743922	308.0	723.0	3768.0	801.0	0.727857	0.298739	0.277728	0.559720	0.268871	...	33.0	28.0	1055.0	284.0	0.777143	0.540984	0.104101	0.668993	0.377772	2
3	0.709192	386.0	849.0	3642.0	723.0	0.719286	0.312551	0.348061	0.586016	0.306753	...	81.0	96.0	987.0	236.0	0.762857	0.457627	0.255520	0.706964	0.407480	3
4	0.674561	455.0	1039.0	3452.0	654.0	0.697679	0.304552	0.410280	0.630258	0.320323	...	147.0	171.0	912.0	170.0	0.756429	0.462264	0.463722	0.732552	0.424788	4
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
195	0.499122	767.0	870.0	3621.0	342.0	0.783571	0.468540	0.691614	0.833873	0.609528	...	222.0	193.0	890.0	95.0	0.794286	0.534940	0.700315	0.848122	0.684006	195
196	0.509964	775.0	858.0	3633.0	334.0	0.787143	0.474587	0.698828	0.827193	0.586164	...	220.0	194.0	889.0	97.0	0.792143	0.531401	0.694006	0.846244	0.680606	196
197	0.501565	777.0	838.0	3653.0	332.0	0.791071	0.481115	0.700631	0.833847	0.609951	...	221.0	198.0	885.0	96.0	0.790000	0.527446	0.697161	0.846367	0.681103	197
198	0.504158	773.0	844.0	3647.0	336.0	0.789286	0.478046	0.697024	0.830421	0.604810	...	220.0	198.0	885.0	97.0	0.789286	0.526316	0.694006	0.845655	0.678282	198
199	0.510945	757.0	845.0	3646.0	352.0	0.786250	0.472534	0.682597	0.826835	0.579900	...	221.0	199.0	884.0	96.0	0.789286	0.526190	0.697161	0.845807	0.679480	199

	model1	model2
loss	0.346845	0.470459
tp	296.000000	440.000000
fp	103.000000	452.000000
tn	2286.000000	1937.000000
fn	315.000000	171.000000
accuracy	0.860667	0.792333
precision	0.741855	0.493274
recall	0.484452	0.720131
auc	0.851803	0.849523
prc	0.685067	0.670289

	loss	tp	fp	tn	fn	accuracy	precision	recall	auc	prc	...	val_tp	val_fp	val_tn	val_fn	val_accuracy	val_precision	val_recall	val_auc	val_prc	epoch
0	0.852494	577.0	845.0	6035.0	1143.0	0.768837	0.405767	0.335465	0.641184	0.410032	...	3.0	1.0	1082.0	314.0	0.775000	0.750000	0.009464	0.663993	0.372692	0
1	0.734511	415.0	1198.0	3293.0	694.0	0.662143	0.257285	0.374211	0.570766	0.257065	...	115.0	122.0	961.0	202.0	0.768571	0.485232	0.362776	0.733171	0.418219	1
2	0.689148	518.0	1240.0	3251.0	591.0	0.673036	0.294653	0.467087	0.627775	0.307522	...	106.0	94.0	989.0	211.0	0.782143	0.530000	0.334385	0.751776	0.422954	2
3	0.668531	490.0	951.0	3540.0	619.0	0.719643	0.340042	0.441839	0.651068	0.330337	...	112.0	91.0	992.0	205.0	0.788571	0.551724	0.353312	0.756898	0.424010	3
4	0.653546	479.0	863.0	3628.0	630.0	0.733393	0.356930	0.431921	0.674920	0.355645	...	119.0	118.0	965.0	198.0	0.774286	0.502110	0.375394	0.758337	0.427406	4
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
195	0.480674	812.0	812.0	3679.0	297.0	0.801964	0.500000	0.732191	0.848983	0.646198	...	222.0	175.0	908.0	95.0	0.807143	0.559194	0.700315	0.847312	0.688487	195
196	0.467797	833.0	840.0	3651.0	276.0	0.800714	0.497908	0.751127	0.858812	0.657732	...	221.0	174.0	909.0	96.0	0.807143	0.559494	0.697161	0.846964	0.687874	196
197	0.478710	807.0	783.0	3708.0	302.0	0.806250	0.507547	0.727683	0.854230	0.633628	...	219.0	171.0	912.0	98.0	0.807857	0.561538	0.690852	0.845197	0.684279	197
198	0.479745	778.0	781.0	3710.0	331.0	0.801429	0.499038	0.701533	0.851853	0.634848	...	216.0	169.0	914.0	101.0	0.807143	0.561039	0.681388	0.846209	0.686232	198
199	0.472510	804.0	788.0	3703.0	305.0	0.804821	0.505025	0.724977	0.857057	0.646645	...	221.0	175.0	908.0	96.0	0.806429	0.558081	0.697161	0.847946	0.689236	199